1991 Intelligent chip control and tool wear estimation in ...
237
University of Wollongong Research Online University of Wollongong esis Collection University of Wollongong esis Collections 1991 Intelligent chip control and tool wear estimation in automated machining systems Xiang Dong Fang University of Wollongong Research Online is the open access institutional repository for the University of Wollongong. For further information contact the UOW Library: [email protected] Recommended Citation Fang, Xiang Dong, Intelligent chip control and tool wear estimation in automated machining systems, Doctor of Philosophy thesis, Department of Mechanical Engineering, University of Wollongong, 1991. hp://ro.uow.edu.au/theses/1585
Transcript of 1991 Intelligent chip control and tool wear estimation in ...
University of WollongongResearch Online
University of Wollongong Thesis Collection University of Wollongong Thesis Collections
1991
Intelligent chip control and tool wear estimation inautomated machining systemsXiang Dong FangUniversity of Wollongong
Research Online is the open access institutional repository for theUniversity of Wollongong. For further information contact the UOWLibrary: [email protected]
Recommended CitationFang, Xiang Dong, Intelligent chip control and tool wear estimation in automated machining systems, Doctor of Philosophy thesis,Department of Mechanical Engineering, University of Wollongong, 1991. http://ro.uow.edu.au/theses/1585
1 / FEB 1992
ERRATA TABLE
Page
32 120 135 141 176 179 190
Line
5 5 from foot 2 from foot 3 from foot
11 11 11
Corrections
was
we form much a
grooves' spacing This work
to modify tool Reconnition
to be corrected as >j
was from much
the spacing of the grooves This work is
to modify the tool Recognition
INTELLIGENT CHIP CONTROL AND TOOL WEAR ESTIMATION IN AUTOMATED MACHINING SYSTEMS
A thesis submitted in fulfilment of the
requirements for the award of the degree
DOCTOR OF PHILOSOPHY
from
THE UNIVERSITY OF WOLLONGONG
by
XIANG DONG FANG
B.E. (Hons.), M.E. (Hons.), Tsinghua
Department of Mechanical Engineering
June, 1991
DECLARATION
This is to certify that the work presented in this thesis was carried out by the author in
the Department of Mechanical Engineering of the University of Wollongong, Australia
and has not been submitted for a degree to any other university or institution.
Xiang Dong FANG
i i
ACKNOWLEDGEMENTS
The author is gready indebted to his supervisors Dr. Y. Yao and Professor G. Amdt
for their excellent guidance, close supervision and generous help during the course of
this thesis.
The author is also deeply grateful to the invaluable guidance and help from his former
supervisor, Associate Professor I.S. Jawahir, currentiy in the University of Kentucky,
U.S.A.
The author is very thankful to the University of Wollongong as well as CAMIA, the
Centre for Advanced Manufacturing and Industrial Automation, for providing him
with postgraduate scholarships.
Acknowledgement is also given to all the staff in the Department, in particular to
Professor P. C. Arnold for bis encouragement and all sorts of help; Dr. E. Siores for
his useful advice on artificial intelligence techniques; Mr. D. Jamieson, Mr. I. Kirby,
Mr. T. Kent and M r K. Maywald for their assistance in using computer and instrument
facilities.
Many thanks are also due to the workshop staff, especially Mr. M. Morillas and Mr.
R. Marshall, for their many days with the assistance to the machining experiments.
Finally, the author expresses his heartfelt thanks to his wife Na Lin for her consistent
encouragement and help during his P h D study.
ABSTRACT
Chip control and tool wear estimation are two major concerns in automated machining
systems. Chip control is essential for the safety of the machining operation, the
maintenance of good surface finish on the machined part, the convenience of chip
disposal, and possible power reduction; while tool wear estimation is vital to an
effective tool change policy and quality control strategy, especially in finish-
machining.
This thesis first presents a new method for quantifying chip breaking and chip shapes
with a fuzzy rating system, and further for predicting the chip breakability for arbitrary
combinations of machining conditions through a fuzzy-set mathematical model. A
predictive expert system for off-line assessment of machining performance, with chip
control as a major criterion and due consideration to surface finish and power
consumption, is then developed.
A knowledge-based system for designing optimum chip breakers is set up with a
criterion of efficient chip breaking at reduced power consumption. The method is
based on the analysis of three-dimensional chip flow in oblique machining for a wide
range of work materials, cutting conditions, tool geometries, chip breaker styles/sizes
and restricted contact lengths.
Experimental results of tool wear patterns in finish-machining show that estimation of
more than one type of tool wear is required to assure the quality of a finished product.
In order to achieve this, a dispersion analysis algorithm, derived from the established
i v
multivariate time series models, is used for the overall estimation of tool wear,
including major flank wear, crater wear, minor flank wear and groove wear at the
minor cutting edge.
Finally, neural network techniques are used for modelling the dynamic
interrelationship between the chip forming behaviour and different tool wear states.
By integrating the developed methods for predicting chip breakability/shapes and for
estimating comprehensive tool wear, the initially-predicted chip forming/breaking
patterns can be updated with tool wear progression during the machining process
through the use of neural network techniques.
The results show that the methods developed in this thesis, for predicting chip
breaking and shapes, and for evaluating machining performance, including chip
control, surface finish and power consumption, may be used to form a basis for the
off-line assessment of "total machinability" for automated machining systems. The
strategy of comprehensive tool wear estimation provides a feasible means for on-line
tool wear monitoring to assure product quality, especially under finish-machining
conditions. The results also show that by using neural networks, chip forming
behaviour with tool wear progression can be evaluated in-process, thus providing a
feasible approach for achieving the on-line assessment of machining performance
including chip forming patterns, surface finish and overall tool wear progression.
TABLE OF CONTENTS
DECLARATION
ACKNOWLEDGEMENTS
ABSTRACT
TABLE OF CONTENTS
LIST OF TABLES
LIST OF FIGURES
NOMENCLATURE
PUBLICATIONS WHILE STUDYING FOR PHD COURSE
Chapter Title
1. INTRODUCTION
1.1 The Importance of the Current Research Project and
Its Objective
1.1.1 Significance of Chip Control in Automated Machining
1.1.2 The Need for Developing a Knowledge-based Expert System for Effective Chip Control
1.13 Strategies of Tool Wear Estimation in Finish-Machining
1.1.4 Multivariate Time Series Modelling for Comprehensive Tool Wear Estimation
1.1.5 Integrating Chip Control and Tool Wear Estimation for On-line Assessment of Machining Performance
1.2 Scope of Thesis
VI
1.3 Literature Survey
PREDICTING CHIP B R E A K A B U J T Y IN MACHINING
W I T H A F U Z Z Y SET-BASED EXPERT SYSTEM 10.
2.1 Introduction 10.
2.2 Fuzzy Description of Chip Breakability 12.
2.3 Machining Experiments for Setting up the Basic Chip Database 15.
2.4 Factors Affecting Chip Breakability: An Experimental Analysis 17.
2.4.1 Effect of Cutting Conditions 20.
2.4.2 Effect of Work Materials 24.
2.4.3 Effect of Tool Inserts 29.
2.4.4 Effect of Tool Geometries 31.
2.5 Foundation of a Fuzzy-set Mathematical Model 33.
2.6 Establishment of an Expert System for Predicting Chip Breakability 37.
2.6.1 Knowledge Representation 37.
2.6.2 Examples of Knowledge Rules for Chip Breakability 39.
2.6.3 Schematic Diagram of the Expert System 44.
2.7 Prolog Programming Architecture 45.
2.8 Summary and Concluding Remarks 46.
A PREDICTIVE EXPERT SYSTEM FOR THE ASSESSMENT
OF MACHINING P E R F O R M A N C E W I T H CHIP C O N T R O L
AS A M A J O R CRITERION 48.
3.1 Introduction 48.
3.2 Foundation of Machining Performance Database 49.
VII
3.3 Assessment of Machining Performance 50.
3.3.1 Assessment of Chip Control 50.
3.3.2 Assessment of Surface Finish 5 3.
3.3.3 Assessment of Power Consumption 5 3.
3.4 An Investigation of the Influencing Factors on Machining
Performance 54.
3.4.1 Effect of Tool Insert Types 54.
3.4.2 Effect of Work Materials 56.
3.4.3 Effects of Cutting Conditions and Tool Geometries 56.
3.5 Evaluation of Work Material Properties 5 8.
3.6 Knowledge Representation 60.
3.6.1 Basic Facts Based on the Databa.se 60.
3.6.2 Knowledge from Expert's Experience 61.
3.7 Establishment of a Predictive Expert System 64.
3.8 Summary and Concluding Remarks 65.
A KNOWLEDGE-BASED SYSTEM FOR DESIGNING
EFFECTIVE G R O O V E D CHIP B R E A K E R S B A S E D O N
THREE-DIMENSIONAL CHIP F L O W IN M A C H I N I N G 67.
4.1 Introduction 67.
4.2 Present Knowledge of Chip Control with Grooved Chip Breakers 68.
4.2.1 Tool Restricted Contact Effect on Chip Breaking 68.
4.2.2 Three-dimensional Chip Flow 69.
4.2.3 Power Consumption Rate 70.
4.3 Establishment of a Database System 71.
4.3.1 Reference Database 71.
v i i i
4.3.2 Grooved Chip Breaker Database 71.
4.3.3 Three-dirnensional Chip Flow Database 72.
4.4 Summary of Present Knowledge for Setting up a Knowledge Base 72.
4.4.1 Analysis of Chip Breaking 73.
4.4.2 Analysis of Grooved Chip Breakers and Power Consumption Rate 80.
4.5 Analysis of 3-D Chip Flow and Chip Curling
with Grooved Chip Breakers 8 3.
4.5.1 Effect of the Chip Backflow Angle 83.
4.5.2 Effect of the Chip Sideflow Angle 85.
4.6 The Establishment of a Knowledge Base 86.
4.6.1 Knowledge Rules for Predicting the Natural Contact Length 86.
4.6.2 Knowledge Rules for Predicting the Chip Backflow Angle 88.
4.6.3 Knowledge Rules for Predicting the Chip Sideflow Angle 90.
4.6.4 Knowledge Rules for Deterrnining the Minimum Power Consumption 92.
4.6.5 Knowledge Rules for Selecting the Effective Groove Profile 93.
4.7 A Strategy for Designing Effective Grooved Chip Breakers 94.
4.7.1 Deterrnining the Optimum Restricted Contact Length 94.
4.7.2 Deterrnining the Effective Groove Parameters 96.
4.7.3 Schematic Diagram for Designing Effective Grooved Chip Breakers 97.
4.8 Summary and Concluding Remarks 99.
5. C O M P R E H E N S I V E T O O L W E A R ESTIMATION IN FTNISH-
M A C H I N I N G B Y MULTIVARIATE TIME SERIES ANALYSIS 100.
5.1 Introduction 100.
5.2 Tool Wear Experiments 102.
5.2.1 Description of Experiments 102.
5.2.2 Definition of Comprehensive Tool Wear Parameters 105.
5.2.3 Tool Wear Measurement and Wear Development Patterns 106.
5.3 Multivariate Time Series Techniques 110.
5.3.1 Trivariate A R M A V Models 110.
5.3.2 Dispersion Analysis 112.
5.4 Tool Wear Estimation by Dispersion Analysis based on A R V Models 114.
5.4.1 A R V Modelling of 3-D Dynamic Cutting Forces 114.
5.4.2 Patterns of Dispersion Development 115.
5.4.3 Analysis Associated with Physical Interpretation 119.
5.4.4 Strategies of Tool Wear Estimation in Finish-Machining 121.
5.4.5 Structural Dynamics and Idle Disturbances 122.
5.5 Summary and Concluding Remarks 122.
M O N I T O R I N G G R O O V E W E A R D E V E L O P M E N T IN FINISH-
M A C H I N I N G B Y A R V M O D E L - B A S E D MULTIPLE DISPERSION
ANALYSIS 124.
6.1 Introduction 124.
6.2 Investigation in to the Patterns of Groove Wear Development 126.
6.2.1 Tool Wear Experiments 126.
6.2.2 Development of Groove Wear 127.
6.2.3 Patterns of Groove Wear Development 131.
6.3 ARV Model-based Multiple Dispersion Analysis 135.
6.3.1 Application of Autoregressive Vector Time 135. Series Modelling
6.3.2 Multiple Dispersion Analysis 136.
6.4 Strategy for Groove Wear Monitoring by Multiple Dispersion Analysis 138.
6.4.1 Groove Wear Monitoring for the Lower Feed Condition 138.
6.4.2 Groove Wear Monitoring for the High Feed Condition 140.
6.5 Physical Interpretations 141.
6.6 Summary and Concluding Remarks 143.
INTEGRATING CHIP C O N T R O L A N D T O O L W E A R
T H R O U G H N E U R A L N E T W O R K S F O R T H E ON-LINE
A S S E S S M E N T O F M A C H I N I N G P E R F O R M A N C E 145.
7.1 Introduction 145.
7.2 Description of Machining Experiments 147.
7.2.1 Machining Conditions 147.
7.2.2 Comprehensive Tool Wear Patterns 148.
7.2.3 Chip Breakability Assessment 149.
7.2.4 Surface Finish Assessment 152.
7.3 Feature Extraction from 3-D Dynamic Cutting Forces 153.
7.4 Architecture of Neural Networks 155.
7.4.1 Neural Network Techniques Used 155.
7.4.2 Selection of Input Features 158.
7.5 Analysis of Results from Neural Networks 161.
7.5.1 Training Strategy with Neural Networks 161.
7.5.2 Analysis of Results for Training and Testing Effects 162.
XI
7.6 Extension: Future Directions and Developments 169.
7.6.1 Real-time Application of Chip Control and Tool Wear Estimation in Automated Machining Systems 169.
7.6.2 Developing a Self-organising Neural Network 171.
7.7 Summary and Concluding Remarks 171.
8. C O N C L U S I O N S 173.
8.1 A N e w Approach for the Quantitative Prediction of Chip Breakability and Chip Forming Patterns 174.
8.2 A Scientifically-based Method for the Effective Design of Chip Breakers 175.
8.3 A N e w Strategy for Comprehensive Tool Wear Estimation 176.
8.4 A N e w Attempt at Estimating Chip Forming Patterns with Tool Wear Progression for On-line Assessment of Machining Performance 177.
8.5 Suggestion for Future Work 178.
REFERENCES 181.
APPENDK
A Standard Fuzzy Membership Values for Chip Breakability
B Representative Data of Surface Finish when Machining
with Different Tool Chip Breakers
C Overall Groove Wear Data from Groove Wear Experiments
D Input/Output Training and Testing Data in Neural Network Experiments
XII
LIST OF TABLES
TABLE PAGE
1.1 ARMA modelling of machining process for condition monitoring 5.
2.1 Membership values for most common chip shapes/sizes 15.
2.2 Machining conditions used for setting up the basic chip database 16.
2.3 Conditions used for the extensive machining experiments 19.
2.4 Fuzzy linguistic values 35.
2.5 Predicting chip breakability with fuzzy ratings 36.
3.1 Chip acceptable factor g(x) in automated machining systems 51.
3.2 Integrated assessment for chip control in automated machining systems 52.
3.3 A n example of work material effect table 5 8.
3.4 Fuzzy ratings of surface finish and power consumption 59.
5.1 Machining conditions used in tool wear experiments 103.
5.2 Tool wear measurement result for cutting condition Group 4 106.
5.3 Tool wear measurement results for cutting condition Groups 1-3 107.
5.4 Dispersion analysis results 116.
5.5 Tool wear analysis for cutting condition Group 1 121.
6.1 Machining conditions used in groove wear experiments 127.
6.2 Relationship between surface roughness and dispersion pattern 141.
6.3 Comparison between dispersion analysis results and physical quantities 143.
7.1 Machining conditions used in neural network experiments 148.
7.2 Chip forming behaviour in machining process while tool wear 150.
7.3 The training data under Training Cutting Conditions 1-3 160.
x i i i
LIST OF FIGURES
FIGURE PAGE
2.1 Classification of chip shapes 15.
2.2 T w o examples of chip breakability diagrams 18.
2.3 Factors affecting chip breakability 18.
2.4 The effect of depth of cut on chip breakability 21.
2.5 The effect of feed on chip breakability 22.
2.6 The effect of cutting speed on chip breakability 23.
2.7 The effect of carbon content on chip breakability 25.
2.8 The effect of Cr-Mo alloying on chip breakability 26.
2.9 The effect of low-carbon-based alloying on chip breakability 27.
2.10 The effect of different work materials on chip breakability 2 8.
2.11 Chip breakability diagrams for six different work materials 28.
2.12 Chip breakability diagram for different tool inserts at a high cutting speed 29.
2.13 Chip breakability diagrams for different tool inserts at a low cutting speed 30.
2.14 Comparison of cutting speed effect with different chip breakers 31.
2.15 The effect of cutting edge angle on chip breakability 32.
2.16 The effect of tool nose radius on chip breakability 33.
2.17 Fuzzy exponential functions 3 5.
2.18 The expert system schematic diagram 44.
2.19 P R O L O G program structure diagram 46.
3.1 Surface finish comparison for four tool inserts 54.
3.2 Power consumption comparison for five tool inserts 55.
3.3 Surface finish comparison for four work materials 56.
3.4 Power consumption comparison for four work materials 57.
3.5 Integrated effects of cutting conditions and tool geometries 57.
3.6 Flow chart for selecting efficient feed value 63.
3.7 A schematic diagram for setting up the predictive expert system 64.
3.8 The predictive expert system acting as an auxiliary for A P P system 65.
4.1 Three chip streaming models 69.
4.2 Three dimensional chip flow with a grooved chip breaker 70.
4.3 Configurations of grooved chip breakers 72.
4.4 Typical relationship diagrams for restricted contact tools 74.
4.5 The effect of tool restricted contact length on chip up-curl and chip breaking 7 5.
4.6 The effect of tool restricted contact length on chip up-curl radius for three chip-breaker groove sizes 7 5.
4.7 The effect of chip-breaker groove style on chip breaking 76.
4.8 The effects of depth of cut 77.
4.9 The effects of feed 78.
4.10 The effects of tool nose radius 79.
4.11 The effect of inclination angle on chip sideflow 80.
4.12 The effect of cutting speed on chip sideflow 80.
4.13 The effects of work material (carbon steels) 81.
4.14 The effect of tool/chip contact length on power consumption 82.
4.15 The effect of tool restricted contact length on specific cutting pressure in machining with grooved chip breakers 8 3.
4.16 The chip backflow effect on tool groove utilisation 84.
4.17 The chip sideflow effect on the effective tool groove parameters 8 5.
4.18 Determination of tool/chip natural contact length 86.
4.19 Determining the optimum restricted contact length 95.
4.20 Flow chart for the knowledge-based system developed 98.
5.1 Experiment setup 102.
5.2 Dynamic cutting force measured in terms of its three orthogonal components 104.
XV
5.3 Definition of comprehensive tool wear parameters 105.
5.4 Tool wear development for cutting condition Group 1 108.
5.5 Tool wear development f or cutting condition Groups 2 & 3 109.
5.6 Tool wear observed under a scanning electron microscope (SEM) 110.
5.7 Dispersion diagram for cutting condition Group 1 117.
5.8 Dispersion diagrams for cutting condition Groups 2 & 3 118.
5.9 The model for the forces acting on different tool faces 119.
6.1 Groove wear at the minor cutting edge 125.
6.2 Number of grooves at the minor cutting edge 128.
6.3 Development of groove depth 128.
6.4 Development of the maximum groove length 129.
6.5 Development of groove wear area 129.
6.6 Development of nose wear 130.
6.7 Surface roughness for five cutting conditions 130.
6.8 Groove wear development under the lower feed condition 132.
6.9 Groove wear development under the higher feed condition 133.
6.10 Dispersion patterns for the lower feed condition 139.
6.11 Dispersion patterns for the higher feed condition 140.
7.1 Development patterns of comprehensive tool wear 149.
7.2 Change of chip shapes/sizes with tool wear progression 150.
7.3 Change of chip breakability under Training Cutting Conditions 2-6 151.
7.4 Change of surface finish with tool wear rates for Training Cutting Condition 1 152.
7.5 Development of surface finish for Training Cutting Conditions 2-6 153.
7.6 A typical relationship between tool wear and dispersion patterns 154.
7.7 Back-propagation neural network with one hidden layer 156.
7.8 The function of artificial neuron 157.
7.9 Description of three basic transfer functions 158.
7.10 Feature selection for neural network 159.
7.11 Effect of the number of hidden neurons for three transfer functions 163.
7.12 Evaluation of training effect of chip breakability 164.
7.13 Evaluation of training effect of surface finish 166.
7.14 Neural network performance in predicting patterns of chip breakability 167.
7.15 Neural network performance in predicting patterns of surface finish 168.
7.16 Proposed schematic diagram of an integrated system for chip control and tool wear estimation based on the G 2 Real-time Expert System 170.
8.1 A proposed .schematic diagram for an expert system-supported tool wear estimation system 180.
xvii
NOMENCLATURE
at white noise
B groove width of tool chip breaker
B e equivalent groove width of tool chip breaker
B j bias of the jfh neuron in a neural network
Cs cutting edge angle
d depth of cut or groove depth of tool chip breaker
d i raised height of groove back wall
d2 reduced height of groove back wall
di dispersion
Dj dispersion percentage
Dxy cross-dispersion between x and y series signals
Dxz cross-dispersion between x and z series signals
Dyz cross-dispersion between y and z series signals
D'x(LF) rate of low frequency (LF) dispersion of the feed force (in x direction)
D'x(HF) rate of high frequency (HF) dispersion of the feed force (in x direction)
D'y(HF) rate of high frequency (HF) dispersion of the thrust force (in y direction)
D'z(HF) rate of high frequency (HF) dispersion of the main cutting force (in z direction)
f feed
fn frequency corresponding to a pair of complex conjugate eigenvalues
F c main cutting force
Ft thrust cutting force
F x cutting force in feed (x) direction or feed cutting force
F y cutting force in thrust (y) direction or thrust cutting force
Fz
F a
Fan
Fph
Fpn
Fpv
F T
Fyn
g(x)
G k
G k
h
he
he
hn
h P
N i
KB
KK
KS
KT
N
Oj
Ps
Fse
q
qit—» qs
cutting force in main cutting (z) direction or main cutting force
tangential (friction) force to major flank
normal force to major flank
horizontally tangential (friction) force to minor flank
normal force to minor flank
vertically tangential (friction) force to minor flank
tangential (friction) force to crater face
normal force to crater face
chip acceptability factor
explicit Green's function
the complex conjugate of G k
tool restricted contact length
tool/chip contact length
effective tool restricted contact length
tool/chip natural contact length
tool restricted contact length with minimum power consumption
tool restricted contact length with effective chip breaking
tool inclination angle
crater width on tool rake face
crater length on tool rake face
retract of tool cutting edge
crater depth on tool rake face
tool nose wear
output pattern of a neural network
specific power consumption
efficient specific power consumption
exponent decided by fuzzy linguistic values
exponents decided by fuzzy linguistic values (corresponding to each machining parameter)
r tool nose radius
Ra arithmetic mean deviation of surface roughness
R k standard chip breakability matrix
Ro groove radius of tool chip breaker
ru chip up-curl radius
rui radius of initial chip curvature
ruf radius of final chip curvature
ti undeforrned chip thickness
t2 chip thickness
Ti eigenvector
Tj target pattern of a neural network
u exponent for determining the variations of chip breakability membership values due to the change of tool geometry
v exponent for determining the variations of surface finish membership values due to the change of tool geometry
V cutting speed
V c chip velocity
V B major flank wear
V B ' minor flank wear
V G length of the groove (notch)
w exponent for determining the variations of power consumption membership values due to the change of tool geometry
or width of cut in orthogonal cutting
Wji weight between the/th neuron and its ith input in a neural network
W L work material evaluation matrix
x chip shape/type
X chip shape set
Xi the ith input of a neural network
X t discrete series of observation vectors
Yj output of the ;th neuron activated by a non-linear transfer function
XX
Zj linear summation of the total inputs to theyth neuron
a tool rake angle
y0 process variation
Yo-vec vector auto-covariance matrix
y k correlation matrix of measured variables
b\ Kronecker delta function
A uniform sample interval
Ah ny increment of tool/chip natural contact length due to the variations of cutting speed
A h n a increment of tool/chip natural contact length due to the variations of tool rake angle
A P n increment of power consumption
AT] b h increment of chip backflow angle due to the variations of tool restricted contact length
Arjbf increment of chip backflow angle due to the variations of feed
Arj b V increment of chip backflow angle due to the variations of cutting speed
A r ^ increment of chip backflow angle due to the variations of tool rake angle
Ar|sd increment of chip sideflow angle due to the variations of depth of cut
AT|sf increment of chip sideflow angle due to the variations of feed
At|sr increment of chip sideflow angle due to the variations of tool nose radius
Eb ultimate strain of the chip material
8 groove tangent angle of tool chip breaker
0j moving average parameter matrix
Xj eigenvalue
p(x) fuzzy membership value
p*(x) modified fuzzy membership value
Pi(k) input of fuzzy membership values for chip breakability to the neural networks used
pk(i, j) fuzzy membership value of the standard chip breakability matrix
Mk)
Up
Us
<*a
^
<*>i
Tib
^ b std
Tlbe
Tls
Tlsi
Tls std
output of fuzzy membership values for chip breakability from the neural networks used
membership value of power consumption
membership value of surface finish
covariance of white noise at
residual matrix
autoregressive parameter matrix
chip backflow angle
standard chip backflow angle in the 3-D chip flow database
effective chip backflow angle
chip sideflow angle
increment of chip sideflow angle due to the variations of tool inclination angle compared with that at the standard 0 degree
standard chip sideflow angle in the 3-D chip flow database
xxii
PUBLICATIONS WHILE STUDYING FOR PHD COURSE
I. Journal Papers
1. Y. Yao, X. D. Fang and G. Arndt, "Comprehensive Tool Wear Estimation in
Finish-machining via Multivariate Time-Series Analysis of 3-D cutting Forces",
Annals of CIRP, Vol. 39(1), 1990, pp. 57-60.
2. X. D. Fang and I. S. Jawahir, "On Predicting Chip Breakability in Machining
of Steels with Carbide Tool Inserts Having Complex Chip Groove
Geometries", to appear in /. Materials Processing Technology, Vol. 26, Issue
2, selected from the papers in Proc. 7th International Conference on Computer-
Aided Production Engineering, Tennessee, USA, August 13-14,1991, pp. 37-
48.
3. Y. Yao, X. D. Fang and G. Arndt, "On-line Estimation of Groove Wear in the
Minor Cutting Edge for Finish-machining", to appear in Annals of CIRP, Vol.
40(1), 1991.
4. Y. Yao and X.D. Fang. "Modelling of Multivariate Time Series for Tool Wear
Estimation in Finish-Turning", paper accepted for publication in Int. J. Machine
Tools & Manufacturing.
II. Papers in International Conference Proceedings
5. I. S. Jawahir and X. D. Fang. "A Knowledge-based Approach for Improved
Performance with Grooved Chip Breakers in Metal Machining", Proc. 3rd
International Conference on Advances in Manufacturing Technology, Organised
by IProdE, Singapore, August 14-16, 1989, pp. 130-144.
6. X. D. Fang and I. S. Jawahir, "An Expert System Based on A Fuzzy
Mathematical Model for Chip Breakability Assessments in Automated
Machining", Proc. Int. Conf, ASME, MI'90, Vol. TV, Adanta, Georgia, U.
S. A., March 25-28, 1990, pp. 31-37.
Y. Yao, I. S. Jawahir, D. Jamieson and X. D. Fang. "Computer Animation of
Chip Flow and Chip Curl in an Expert Process Planning System for Metal
Machining", Transactions of the North American Manufacturing Research
Conference(NAMRC 1SME), Pennsylvania, USA, May 23-25, 1990, pp. 161-
166.
I. S. Jawahir and X. D. Fang. "Some Guidelines on the Use of a Predictive
Expert System for Chip Control in Automated Process Planning", Proc. Fifth
International Conference on Manufacturing Engineering, Vol.1,Wollongong,
Australia, July 11-13, 1990, pp. 288-292.
Y. Yao, X. D. Fang and R. Rudziejewski, "Analysis of Dynamic Components
of 3-D Cutting Forces in Oblique Machining", Proc. Fifth International
Conference on Manufacturing Engineering, Vol.2,Wollongong, Australia, July
11-13, 1990, pp. 349-353.
X. D. Fang and Y. Yao, "Expert System-Supported On-line Tool Wear
Monitoring", Proc. of International Conference on Artificial Intelligence in
Engineering - AIE'90, Sponsored by I.E.E.(UK), August 21-22, 1990, Kuala
Lumpur, Malaysia, pp. 77-84.
X. D. Fang. Y. Yao and G. Arndt, "Tool Wear Estimation by Multidimensional
Autoregressive Spectral Analysis", Proc. 1990 Pacific Conference on
Manufacturing, Vol. 2, Sydney & Melbourne, Australia, December 17-21,
1990, pp. 834-841.
E. Siores, X. D. Fang and Y. Yao, "Intelligent Design for Manufacturing
Employing Knowledge Based Expert System", accepted for publication in
International Conference on Information Technology for Advanced
Manufacturing Systems, organised by IFTP, Nanjing, P.R. China, Sep., 17-19,
1991.
1
CHAPTER 1
INTRODUCTION
1.1 THE IMPORTANCE OF THE CURRENT RESEARCH
PROJECT AND THE ITS OBJECTIVE
1.1.1 Significance of Chip Control in Automated Machining
It has been long known that chip control in metal machining, particularly in
continuous mode operations such as turning, is vital due to its significant role in
producing small and handleable sized chips for disposal, and in protecting the
machined work surface, cutting tool, machine tool and operator (in manual operation)
from long, snarled, hard and hot unbroken chips. With the advent of automated
manufacturing technologies involving unattended machining operations, the need for
chip control has grown into significant proportions. In order to achieve total chip
control in automated manufacture, the predictability of chip shapes/breaking is
essential. Recently, effective chip control has been recognised by the International
Institution for Production Engineering Research (CIRP) as one of the most urgent
technical problems needed to be solved to improve machining quality [1].
1.1.2 The Need for Developing a Knowledge-based
Expert System for Effective Chip Control
Since the present knowledge of chip control is inadequate to describe quantitatively the
processes of chip flow, chip curl and chip breaking, it becomes very difficult to
predict chip forming behaviour in the actual machining process with a certain degree
3 0009 02980 9915
2
of accuracy. Indeed, the present knowledge of chip control is found to be segmented
and scattered; and currently available machinability database systems have not
incorporated the chip control factor either. Individual factors influencing chip
breakability, such as tool chip breaker configurations, tool geometries, work material
properties and cutting conditions, have not been fully studied to provide a basis for
optimal chip breaking. Although several hundred types of cutting tools with different
toolface configurations involving various chip groove profiles, obstruction lumps,
wavy cutting edges and curved rake faces have become commercially available, most
of these cutting tools are designed on the basis of the traditional "fry and see"
experimental methods. Thus it is apparent that a more scientific methodology is
required to avoid these time-consuming and less-accurate methods.
It is noticed, however, that much of the knowledge and many practical techniques
about machining have not been fully utilised in assessing the chip control effects.
Therefore it is imperative to develop a knowledge-based expert system, integrating
experts' rich experience, empirical rules, experimental results and the existing
qualitative knowledge and theories. The need for developing knowledge-based
systems to address the problems concerning chip control has been highlighted based
on the results of an extensive survey on chip control covering over 160 published
papers [2].
1.1.3 Strategies of Tool Wear Estimation in Finish-Machining
In automated machining, on-line tool wear estimation plays an important role in
establishing an efficient tool change policy and an effective quality control strategy.
Numerous methods have been developed, as reviewed in several survey papers [3-6].
Most of them, however, are concerned with the estimation of major flank wear alone
3
[7-17] or with major flank and crater wear combined [18-21], In actual machining
operations, especially in a finish-machining process, minor flank wear, groove wear
at the minor cutting edge and nose wear are crucial to dimensional accuracy and
surface quality. It is likely that in finish-machining, die wear state at the minor cutting
edge reaches its critical point earlier than those in the major flank and crater, such that
the tool monitoring policy should be established on the basis of the wear state at the
minor cutting edge.
The mechanism of minor flank wear and groove wear at the minor cutting edge is
complex. Minor flank wear will always form in machining process, while under
certain cutting conditions, the grooves, spaced at a distance equal to feed, may appear
and become the dominant type of wear at the minor cutting edge. W h e n no grooves
form at the minor cutting edge, the machined work surface is further shaped by the
minor cutting edge, thus causing minor flank wear mainly due to high-speed friction.
Surface finish and dimensional accuracy are affected by the severity of minor flank
wear.
Although minor flank wear and groove wear at the minor cutting edge have been
recognised long ago to be crucial to the surface quality in finish-machining and much
work has been done on their mechanism [22-26], almost no work has been reported
on the detection and estimation methods for these types of tool wear, presumably due
to the complexity involved in interpreting multivariate signals and resolving them for
estimation of more than one type of wear. Therefore, a more effective monitoring
strategy involving multi-sensor and/or multi-modelling is called for in order to
estimate more than one type of tool wear simultaneously.
4
1.1.4 Multivariate Time Series Modelling for
Comprehensive Tool W e a r Estimation
Autoregressive moving average (ARMA) time series analysis has been used in
modelling machining processes during the past decade. As it does not require much
prior knowledge about the underlying system physics, A R M A modelling is most
suitable for cutting processes which are stochastic in nature. By using A R M A
modelling techniques, a mathematical model which describes the internal system
dynamics can be established solely based on experimentally measured data. Once the
A R M A models are established, feature extraction can be developed in different ways,
such as, residual analysis [27], parametric spectral analysis [28-29], dispersion
analysis [28, 30-32], damping ratio calculation [9, 33], model parameters [7-8, 34],
etc. W h e n more than one series of data is involved, multivariate A R M A vector
models [30, 32, 35-36] may be developed to reveal the interactive characteristics
between different series. Table 1.1 summarises applications of A R M A modelling in
on-line monitoring concerning machining processes.
As seen from Table 1.1, most methods use only the univariate time series model to
estimate tool wear. If more than one quantity is to be estimated, more complexity will
be encountered. Thus a higher demand is placed on effective signal processing and
analysis techniques which shall be able to single out particular signal ingredients
sensitive to particular quantities to be estimated.
5
Table 1.1 A R M A modelling of machining processes for condition monitoring
Refer
ences
[7]
181
[9]
[10]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
Year
1987
1989
1980
1982
1986
1980
1990
1990
1977
1991
1985
1986
Specifications
of
Application
major flank wear
estimation in turning
major flank wear
estimation in turning
major flank wear
estimation in turning
major flank wear
estimation in turning
tool breakage detection
on-line identification
of chatter in turning
estimations of crater wear
and minor flank wear
comprehensive tool
wear estimation
in finish-turning
chatter identification
estimation of groove wear
at the minor cutting edge
estimation of average flank
wear in drilling process
tool breakage detection
in milling process
Signals
Used
acoustic
emission
acoustic
emission
vibration
2-D forces
&
1-D vibration
cutting torque
2-D vibration
&
cutting forces
3-D dynamic
cutting forces
3-D dynamic
cutting forces
vibration
3-D vibration
thrust torque
cutting force
Model
Dimen
sions
l-D
l-D
l-D
l-D
l-D
l-D
3-D
3-D
l-D
3-D
l-D
l-D
Diagnostic Methods
and
Feature Extraction
model parameters
model parameters
damping ratio
actual power
contribution of A R M A
spectral analysis
residual analysis
autoregressive
spectral density &
dispersion analysis
autoregressive
spectral analysis
dispersion analysis
dispersion analysis
multiple (cross)
dispersion analysis
normalised damping ratio
prediction errors based
on model parameters
6
1.1.5 Integrating Chip Control and Tool W e a r Estimation for On-line Assessment of Machining Performance
Chip control and tool wear estimation are two vital yet closely interrelated concerns in
automated machining systems. The dynamic chip forming behaviour will vary
significandy at different tool wear stages, thus resulting in the changeable performance
of the undergoing machining operation, such as chip breaking/chip forms and surface
finish. Although much work has been done on them individually, no work has been
reported to integrate them for on-line monitoring of machining process or optimal
control of machining operation. In fact, all the present machining theories and
machinabihty database are based on unworn or fresh cutting tools. Due to the extreme
complexity involved, it is very difficult at present to model analytically the dynamic
interrelations between the overall machining performance in terms of chip control and
tool wear development
The recent advent of neural network techniques provides a suitable tool for analysing
such a complicated process by extracting knowledge only based on the observations
of experimental input-output data, and further using this knowledge to synthesise an
optimsd predicting model. Some work has been reported on using neural network in
modelling of machining process, such as tool wear monitoring [11-12, 37] and
optimisation of machining process [38].
1.2 SCOPE OF THESIS
The work described in this thesis is devoted to developing an effective means for chip
control and tool wear estimation, two vital concerns in automated machining systems.
The research has been focused on the following four aspects:
7
(a) Predictions of chip breakability for a wide range of machining conditions
through a fuzzy-set mathematical model and further predictions of the
machining performance including chip breakability/chip forming patterns,
surface finish and power consumption.
(b) Development of a knowledge-based system for optimal design of conventional
groove-type chip breaker in the point of view of three-dimensional chip flow in
oblique machining process.
(c) Development of dispersion analysis algorithm, derived from the established
multivariate time series models, for the comprehensive tool wear (major flank
wear, crater wear, minor flank wear and groove wear at the minor cutting edge)
estimation in finish-machining.
(d) Integration of chip control and comprehensive tool wear estimation, with
consideration of surface finish as well, by the use of neural network
techniques, aiming at developing an effective strategy for on-line assessment of
machining performance.
The thesis is organised into 8 chapters, the outline of each chapter is given below.
Chapter 1 highlights the significance of the research project and describes the
objectives to be achieved The outiine of the thesis is also included.
Chapter 2 presents a new approach of a fuzzy rating mechanism to predict the chip
breakability for various chip shapes/sizes produced in the machining process. A
fuzzy-set mathematical model is developed to describe the variations of chip
breakability at different input machining conditions.
8
Chapter 3 is devoted to developing an off-line predictive expert system for the
machining performance involving chip control as well as surface finish and power
consumption based on the extensive machining experiments.
Chapter 4 describes the development of a knowledge-based system for optimal design
of a groove-type chip breaker based on the analysis of three-dimensional chip flow
(sideflow & backflow) and the results of the machining experiments.
Chapter 5 is devoted to investigating the comprehensive tool wear (major flank, crater
& minor flank wear) patterns in finish-machining, and to developing a multivariate
modelling algorithm, aiming at providing an effective estimation strategy which is
able to single out particular signal ingredients sensitive to particular quantities to be
estimated. Dispersion analysis algorithm based on the multivariate A R M A vector
models of 3-D dynamic cutting force is established for extracting features sensitive to
the development of various types of tool wear.
Chapter 6 describes the experimental investigation of the patterns of groove wear at
the minor cutting edge in finish-machining as well as the relationship between groove
formation and surface roughness development Multiple (cross) dispersion analysis is
developed to analyse quantitatively not only the characteristics of individual variables
from 3-D tool vibration produced in the machining process, but also the interactions
among different variables, providing effective detection and estimation of groove wear
at the minor cutting edge.
Chapter 7 aims at integrating chip control and comprehensive tool wear estimation for
on-line assessment of dynamic machining performance, including chip breaking/chip
forms, surface finish and tool wear states. Neural network techniques are used to
9
model the dynamic characteristics of chip forming behaviour as well as surface
roughness development during the process of tool wear.
Chapter 8 presents concluding remarks and summarises the achievements described in
the thesis. Suggestions for future work are also given in this chapter.
1.3 LITERATURE SURVEY
Since each chapter of this thesis is relatively independent, the relevant literature survey
has been integrated into each chapter (from Chapters 1 to 7).
10
CHAPTER 2
PREDICTING CHIP BREAKABILITY IN MACHINING WITH A FUZZY SET-BASED EXPERT SYSTEM
2.1 INTRODUCTION
Chip breaking is essential in metal machining. The advent of automated machining
systems has placed even more emphasis on the need for controlling the chip.
Therefore, developing feasible means for predicting chip breakability and
forms/shapes is highly called for. Due to a great number of interacting process
variables involved, predicting chip breakability has been a major yet complicated
problem for many years. It has been shown, when machining with a conventional
flat-faced tool or a tool with simple chip-breaker configuration, the prediction of chip
breakability can be made with some accuracies under certain machining conditions
[39-41]. However, when tool inserts with complex chip forming toolface
configurations, most of which have been commercially available, are used, predicting
chip breaking becomes much more difficult. This is attributed to the very complex
nature of chip flow mechanics on the toolface, which results in non-uniquely definable
chip curvatures. In this case, it is almost impossible to predict chip breakability with
sufficient accuracy in machining with tools having such complicated chip breaker
configurations. A s such, the need for adopting non-algorithmic means of predicting
chip breakability has been felt by the research workers for quite some time [2] but it is
recently that the problem has been addressed.
Henriksen's [42-43] early works on obstruction-type and integral groove-type chip
breakers laid the foundation for applied research on chip breaking. Following
Colwell's chip flow investigation [44], Stabler [45] established his still famous chip
11
flow law, in which he derived an expression for the direction of chip flow in terms of
the cutting edge inclination angle and showed the validity of this relationship with a
material factor added.
Cook et al [46] explained various chip flow and chip curl mechanisms and showed
that reducing the average friction on the toolface will produce thinner chips. They
also reported that a toolface crater and Built-up Edge (BUE) determine the initial chip
curl which was shown to vary from one material to another. Nakayama, in his over
25 years of experimental and analytical work on metal machining, published a number
of papers on chip breaking [39-40, 47-52]. The most notable early work by
Nakayama [39] in 1962 presented the very first criterion for chip breaking. In this
work he showed that the chip will break when the strain on the chip surface exceeds
the ultimate strain of chip material (eQ, and that chip breakability depends on the chip
thickness te) and the radii of initial and final chip curvature (rui and ruf), in a
combined up and side curling mode, as shown in Equation (2.1).
Worthington et al [41, 53-55] conducted a series of experimental work on chip
breaker geometry and presented their results, which highlight the actual action of a
chip-breaker tool, by showing the effect of a grooved toolface on chip curl and chip
breaking. Johnson [56] reported the applicability of a slip-line field for machining
with restricted contact tools using a velocity diagram (hodograph). Based on this
slip-line field theory, Usui et al [57-58] expanded the basic knowledge on machining
with restricted contact tools. However, the effect of chip streaming (i.e. chip
backflow) which would primarily determine the chip breaker parameters needed for an
effective chip breaking operation, was paid far less attention in research for about 25
1 2
years. This vital effect exists, as shown in recent work [59-62], for tools with a
restricted contact length (on the toolface) less than the tool/chip natural contact length
under a given set of cutting conditions. Subsequendy, Jawahir [2] in an extensive
survey on chip breaking identified the need for knowledge-based systems and
continuous process monitoring for "total chip control" in unmanned machining
systems.
In this chapter, a new approach of fuzzy rating mechanism is presented to quantify the
chip breakability for various chip shapes/sizes produced in the machining processes.
A fuzzy mathematical model was developed to describe the variations of chip
breakability at different input machining conditions. Using the basic representative
experimental results that are stored in the chip database, in conjunction with a set of
appropriate knowledge-rules summarised on the basis of the extensive machining
experiments on chip breaking and chip shapes, the predictability of chip breaking can
be made through the use of an expert system with quite reasonable accuracy.
Knowledge-rules are developed by using Turbo P R O L O G language. The application
of a fuzzy-set mathematical model provides the basis for the establishment of the
expert system.
2.2 FUZZY DESCRIPTION OF CHIP BREAKABILITY
As a result of surveying the development history of predicting chip breakability, it is
apparent that no quantitative methods have been presented on the prediction of chip
breaking and classification of chip shapes although much work has been done on their
analysis [51, 55, 61-63]. In fact, chip breakability is a concept involving many
uncertainties without well-defined boundaries. In industrial practice, when w e explain
chip breakability, descriptive words, such as good, moderate, poor, are usually used.
Therefore, the fuzzy set theory, apparendy very applicable to the problem concerned,
13
is selected to describe quantitatively different levels of chip breakability. In this way,
a description of chip breakability may be expressed in terms of linguistic values based
on the recorded observations. Figure 2.1 shows a detailed classification of chip
shapes produced from machining operations, which w e define as a chip shape set X :
X = { x }, where x is a chip shape/type (2.2)
In order to describe the chip breakability for different chip shapes, we introduce a
fuzzy membership function p(x) within the interval [0,1]:
p(x) e[0,l], x€X (2.3)
In this way, each chip shape is assigned to a fuzzy membership value p(x) which
represents the membership grade of the chip shape x. Taking p(x) = 1.0 as the most
ideally broken chip shape, the membership value p(x{) = 0.7 means that the
breakability rating for chip shape Xj is 0.7. The larger the membership value, the
better the chip breakability. A range of approximate chip breakability ratings
(membership values) has been assigned to various common chip shapes obtainable,
based on the relative ease/difficulty of producing them. Table 2.1 shows the
membership value (representing chip breakability) for each chip shape shown in
Figure 2.1.
1 1
Broken Long
Com. Long
1 A j ^
TUBULAR
CHIPS
1
it SO *3pH
'] I It
CO X
1 °
-<
1 ^
H
i |
\
i t
< Cfl V P.
j 2 X
if
if 60 ^b
if w OS U cfl Cfl O,
^ X a: O O U
r
4«
,-Ua ^
l ^
Cfl vJ
S Mi
« »
(j T3
K 4 4
* *
i - r. •a. m 3
< g
t: 2 '
r
! V
c \
ra ?'
f -- / j -, i
CJ C, *
V
g 2 [3 g
•
3 „jr%
<
1 'IV
" * • %
TOOTHED-EDGE
CHIPS
1 5
Table 2.1 Membership values for most common chip shapes/sizes
Tubular
Chips
Ribbon
Chips
Helical
Chips
Cork
Screw
Chips
Spiral
Chips
Arc
Chips
String
Chips
Tooth-
Edged
Chips
Large
Diameter
0.10
Snarled
0.20
Large
Diameter
0.08
Snarled
0.25
Wavy
0.44-0.48
Up-curl
0.88-0.92
Continuous
Long
0.35
Long
0.28
Snarled
0.28
Long
0.30
Snarled
0.20
Continuous
Long
0.32
Few Turns
0.42-0.48
Side-curl
0.85-0.90
Broken
Long
0.42
Short
0.60
Continuous
Long
0.35
Small Snarled
0.3-0.45
Continuous
Long
0.30
Broken
Long
0.41
Full Turn
0.65-0.67
Connected
0.92-0.95
Medium
0.49
Side-curl
Arc
0.86
Broken
Long
0.43
Broken
Long
0.38
Medium
0.45-0.47
Flat
0.57-0.60
Short
0.64
Connected
Arc
0.92
Medium
0.46-0.48
Medium
0.44-0.46
Short
0.62
Conical
0.67-0.70
Short
0.64
Short
0.60
2.3 MACHINING EXPERIMENTS FOR SETTING UP THE BASIC CHIP DATABASE
A series of experimental work was conducted for setting up the basic chip database to
be used as the primary standard (i.e. reference) for chip breakability. In the basic chip
database, combinations of tool inserts and work materials together with a fairly wide
range of cutting conditions, as shown in Table 2.2, are used.
1 6
Table 2.2 Machining conditions used for setting up the basic chip database
TOOL INSERT
WORK MATERIAL
CUTTING SPEED DEPTH OF CUT FEED
TNMG160408 (ENZ, ENA & ENT) with tool geometry : 0°, 5°, -6°, 90°, 60°, 0.8
CS1020 K1040 EN25
50, 75, 100, 150, 200 rn/min
0.25, 0.5, 1.0, 2.0, 3.0, 4.0 m m
0.06, 0.1, 0.2, 0.3, 0.4, 0.5 mm/rev
Based upon the basic chip database the standard chip breakability matrices Rk's are set
up under the five specified machining parameters (i.e. tool inserts, work materials,
cutting speeds, depths of cut and feeds) as follows :
/
V
uk(l,l) pk(l72) Mkd/6)
uk(2,l) pk(2,2) Mk(2,6)
• • • • • • • • • • • • • • • •
.... pk(ij) .....
• • • • • • • « • • • • • • • •
)*k(6,i> uk<6'2> M*fi) \
\
/
(2.4)
where pjc( i, j) provides the fuzzy membership values of chip breakability, i = 1,
..., 6 and j = 1,..., 6 represent various depths of cut (6 values) and feeds (6 values)
respectively, and k = 1,..., 45 represents combinations of tool inserts (3 values),
work materials (3 values) and cutting speeds (5 values). In effect, 45 individual
matrices were compiled using the experimental results. All these standard fuzzy
membership values for chip breakability are given in Appendix A.
1 7
Figure 2.2 shows two samples of photographs from the machining experiments which
correspond to two of the above standard chip breakability matrices, in the form of
feed-depth of cut relationships at two typical cutting speeds (i.e. at 50 and lOOm/min).
2.4 FACTORS AFFECTING CHIP BREAKABILITY : AN EXPERIMENTAL ANALYSIS
Chip breakability is affected by many factors (Figure 2.3), of which the most
significant factors considered in the present work are:
(a) Cutting conditions (cutting speed, depth of cut and feed)
(b) Work material (7 different work materials used)
(c) Chip breaker type (8 different chip breaker configurations u.sed)
(d) Tool geometry (cutting edge angle and tool nose radius variations considered)
4.0
E E
3
u H-
o •C
** Q.
D
3.0
2.0
1.0
0.5
0.25
£*7TL
,•©
1* *te.
' &
«t>
S*
"i J* *
{x(~-
*Vt t
^
^
•%*—+dS
*****
& %
S x & $
d~t<0
*****
^
5 «P*
** *
X"
0.06 0.10 0.20 0.30
Feed (mm/rev)
0.40 0.50
(a) Cutting Conditions Tool Insert: TNMG160408ENZ Work Material :K1040 Cutting Speed: 50 m/min
18
4.0
E 3.0
o o
2.0
1.0
a. o Q 0.5
0.25
^ft
<L) ^
/
L^-l «J
^
Q»v
^Sy* <
^
^
^ «....
rtwJ
1ST
—
mMl-mm
. > * /
X^ /
/
fl&
££ <y y^
• »%*
« «V * w
^ v
^ '
0.06 0.10 0.20 0.30 0.40
Feed (mm/rev)
0.50
(b) Cutting Conditions : Tool Insert: TNMG160408ENZ Work Material :K1040 Cutting Speed: 100 m/min
Figure 2.2 Two examples of chip breakability diagrams (Feed-Depth of Cut Relationship)
Chip
Breaker Type
Tool Geometry
Cutting Conditions
Tool Material
Work Material
Cutting Fluid
Machining Operation Type (i.e. Turning, Milling, etc.)
Figure 2.3 Factors affecting chip breakability
1 9
The basic chip database is only used as a primary criterion to provide the standard chip
breakability for the machining conditions exactly the .same as those in the basic chip
database. A s chip breakability is a complicated physical phenomenon, influenced by
numerous factors, only the basic chip database established under the selected
machining conditions m a y not be accurate enough to describe the chip breakability
under arbitrary machining conditions/parameters. In order to extend the prediction
range to any combinations of machining conditions, further justification is required.
Table 2.3 Conditions used for the extensive machining experiments
CUTTING
CONDITIONS
OJTTING
TOOLS
WORK
MATERIALS
TOOL
GEOMETRIES
V=30--300 m/min d=0.25--5 mm f=0.06--1.0 mm/rev
1. flat-face (without chip breaker)
2. ENA type: chip breaker with varying widths and wavy cutting edge
3. ENZ type: straight-style grooved chip breaker with raised back wall
4. ENT type: conventional grooved chip breaker
5. ENK type: multiple-lumps chip breaker
6. ENG type: chip breaker configuration combining groove with lump
7. ENJ type: grooved chip breaker with smaller restricted contact length
8. EFJ type: grooved chip breaker with larger restricted contact length
9. EFB type: double-step chip breakers
1. low carbon steel CS1020 : C 0.2% Mn 0.6% (BHN=185)
2. medium carbon steel K1040 : C 0.4% Mn 0.6% (BHN=198)
3. high carbon carbon steel K1055 : C 0.55% Mn 0.6% (BHN=220^
4. Ni-Mo-Cr alloy EN25 : C 0.3% Mn 0.6% Ni 2.5% Mo 0.5% Cr 0.5%
(BHN=240)
5. Mo-Cr alloy AISI 4140 : C 0.4% Mn 0.8% Mo 0.2% Cr 0.9%
(BHN=300)
6. Ni-Cr allov EN36A : C 0.12% Mn 1.0% Ni 3.2% Cr 0.9% (BHN=210)
7. free-machining steel S12L14 : C 0.15% Mn 1.0% S 0.3% Pb 0.25%
(BHN=168)
1. nose radius, r = 0.4, 0.8, 1.2 & 1.6 m m
2. cutting edge angle, Cs = 90, 75, 60 & 45°
20
Therefore, extensive machining experiments, as shown in Table 2.3, were conducted
to analyse the effect of each machining parameter, i.e., work materials, cutting
conditions, tool chip breakers and tool geometries, on chip breakability, aiming at
providing a quantitative reference for determining the interrelationships between chip
breakability and various machining parameters.
Some results of the experimental work and analysis are presented below as
representatives to show the effects of these four factors through the chip breakability
rating (i.e. fuzzy membership value).
2.4.1 Effect of Cutting Conditions
2.4.1-1 Analysis of depth of cut
Three feed values (0.1, 0.2 and 0.3 mm/rev) and three cutting speeds (50, 100 and
200 m/min, representing low, medium and high speeds) were used in the analysis of
chip breakability, represented by membership values p(x), at varying depths of cut.
The effect of depth of cut varies at different feeds. As shown in Figure 2.4 for a
typical E N Z type tool insert when machining work material K1040, at a low feed,
such as O.lmm/rev, the chip breakability decreases with the increase in depth of cut
but the effect is opposite when the feed is increased up to 0.3 mm/rev. The reason for
this is that, in general, as the depth of cut is increased, the chip shapes produced at low
feeds change from long to snarled chips, while at a higher feed (f = 0.3 mm/rev) from
long to small chips. W h e n the depth of cut is larger than 3 m m , in general, there is
virtually no change in chip breakability within the range tested. At lower depths of cut
the chip breakability is generally poor.
21
1.0
0.9-
0.8-
0.7-
0.6-
0.5--
0.4
0.3
0.2-
0.1 -
0.0
<D 3 (0
>
XI
E e
H f=0.1 mm/rev • f=0.2mm/rev o f=0.3mm/rev
-i 1 1 1 r — T 1 (—i 1 1 1 1 1 1 1 1 1 1 , 1 , 1 j —
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 Depth of Cut(mm)
(a) L o w Cutting Speed (50m/min)
0 3 co >
a XI
E
0.0
H f=0.1 mm/rev • f=0.2mm/rev o f=0.3mm/rev
H B B ' - O H H H H
- 1 — I — I — I — I — I — I — I — I — I — I — I — I — I — 1 — I — I — I — I — I — 1 — I — I — I —
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 Depth of Cut(mm)
(b) Medium Cutting Speed (lOOm/min)
o 3 a >
e XI
E e
0.0
H f=0.1 mm/rev • f=0.2mm/rev o f=0.3mm/rev
-i 1 1 1 1 1 1 — T 1 1 r—i 1 1 1 1 ' I ' 1 "—I ' T -0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
Depth of Cut(mm)
(c) High Cutting Speed (200m/min)
Figure 2.4 The effect of depth of cut on chip breakability (Tool Insert: T N M G 1 6 0 4 0 8 E N Z , W o r k Material: K1040)
22
2.4.1-2 Analysis of feed
Chip breakability (p(x)) with varying feeds was estimated at three values of cutting
speeds and two values of depths of cut (see Figure 2.5). Feed plays a very important
role in deterrnining chip breakability. In general, the higher the feed, the better the
chip breakability except at low cutting speeds due to the possible formation of Built-up
Edge (see Figure 2.5). For a larger depth of cut (d = 2 m m ) , w h e n the feed goes up
from 0.2 to 0.3 mm/rev, there is a steep increase in chip breakability, while for a
smaller depth of cut (d = 1 m m ) , only a slight increase being noted. At low depths of
cut, the chip breakability is found to be poor for all feed values.
1.0
0.9
S 0.8 H
0 0.7-3 to
> 0.6 H
"£ 0.5 CD I-
0 •S 0.4-E 0
0.3
0.2
0.1 -i
0.0
Tool Insert: TNMG160408ENZ
Work Material: K1040
0.00 0.20 — I —
0.40
+ X
V=50m/min,d=1mm V=50m/min, d=2mm
V=100m/min,d=1mm
V=100m/min,d=2mm
V=200m/min,d=1mm
V=200m/min,d=2mm — i —
0.80 0.60
Feed(mm/rev)
Figure 2.5 The effect of feed on chip breakability
1.00
2.4.1-3 Analysis of cutting speed
The chip breakability ratings were estimated for three values of depth of cut (1,2 & 4
m m ) and three values of feed (0.1, 0.2 & 0.3 mm/rev) at varying cutting speeds.
Shown in Figure 2.6 is a sample result produced for machining of material K1040
23
CO 3
CO
> o. Jc 0 v. o XI
E 0
a f=o.1 mm/rev
• f=0.2 mm/rev
o f=o.3 mm/rev
100 150 200 Cutting Speed(m/min)
(a) L o w Depth of Cut (1mm)
X
O 3 0 >
JZ CO
0 XI
E 0 2
Q f=0.1 mm/rev • f=0.2 mm/rev o f=o.3 mm/rev
50 100 150 200 Cutting Speed(m/min)
(b) Medium Depth of Cut (2mm)
250 300
x,
0 3 CO
>
JL. 0 w 0 Xl
E 0
H f=o.1 mm/rev • f=0.2 mm/rev o f=o.3 mm/rev
100 150 200 Cutting Speed(m/min)
(c) High Depth of Cut (4mm)
Figure 2.6 The effect of cutting speed on chip breakability
24
with an E N Z tool insert. A s seen, the lower the cutting speed the better the chip
breakability, because higher cutting speeds produce thinner chips which are more
difficult to break (see Equation 2.1). W h e n the cutting speed is lower than 50 m/min,
the chip breakability appears to be very good at all depths of cut for all feed values
tested due to the apparent existence of B U E . W h e n the cutting speed is larger than
50m/min, there are different effects at different depths of cut. For small depth of cut
(Figure 2.6(a)), there is a sharp decrease in chip breakability at all feeds, while for
high depth of cut (see Figure 2.6(c)), significant decrease occurs only at low feeds.
W h e n the cutting speed is beyond 150m/min, there is virtually no change in chip
breakability.
2.4.2 Effect of Work Materials
In the present analysis seven different work materials were considered. The chemical
composition is a major factor which determines the mechanical properties of a work
material, this in turn affects the chip breakability. Several chip breakability diagrams
for conditions using different work materials were plotted and some of the
representative diagrams are shown in Figures 2.7 to 2.11.
2.4.2-1 Analysis of carbon content on chip breakability
Three carbon steels, with the same chemical composition but different carbon
contents, were selected to investigate the influence of carbon content on chip
breakability. In Figures 2.7(a), (b) and (c), the change of membership values was
plotted with varying cutting speeds, feeds and depths of cut respectively. The results
show that the relationships among different carbon contents are consistent in all the
cases, i.e. a steel with higher carbon content is more difficult to break, thus with
smaller membership value.
25
x_
0 3 a >
a XI
E o
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
Q CS1020 + K1040 o K1055
d=2 mm - s=0.2mm/rev ENT Insert
a CS1020 + K1040
L o K1055
d=2 mm V=100m/min ENT Insert
T
s=0.2mm/rev V=100m/min ENT Insert
T -«—r 0 100 200 0 0.20 0.40 0.600 1.0 2.0 3.0 4.0 5.0 Cutting Speed (m/min) Feed (mm/rev) Depth of Cut (mm)
(a) Varying Cutting Speed (b) Varying Feed (c) Varying Depth of Cut
Figure 2.7 The effect of carbon content on chip breakability
2.4.2-2 Analysis of Cr-Mo alloy steel
The work material AISI4140 is a typical Cr-Mo alloy steel widely used in industrial
practice. In order to study the effect of Cr-Mo alloy element, a comparison was made
with plain carbon steel, K1040, both having the same carbon content. As shown in
Figure 2.8, there are generally no significant difference in chip breakability at low (0.1
mm/rev) and high (0.3 rnm/rev) feeds because the chip breakability of both materials is
poor at low feeds and good at high feeds with the selected cutting speed (200 m/min).
However, when the feed is at its medium value (0.2 mm/rev), there is an abrupt increase
in chip breakability for material K1040, while for material AISI4140, the change occurs
only when the feed goes up to 0.3 mm/rev, due to its higher chip ultimate strain than that
of material K1040.
26
2.0 3.0 Depth of Cut (mm)
Material AISI4140
o f=o.1 mm/rev • f=0.2mm/rev
+ f=0.3mm/rev
Material K1040
• f=0.1 mm/rev x f=0.2mm/rev a f=0.3mm/rev
TNMG160408ENZ V = 200 m/min
5.0
Figure 2.8 The effect of Cr-Mo alloying on chip breakability
2.4.2-3 Comparative analysis for three low-carbon-based alloy steels
The membership values of three work materials with approximately equal low carbon
content yet different alloy elements were shown in Figure 2.9 at varying depths of cut.
Generally speaking for all three materials, with the increase of depth of cut (up to 3 mm),
the chip breakability increases at the higher feed (0.2 mm/rev) while decreases at the
lower feed (0.1 mm/rev). Among the three work materials, it is clear that the chip
breakability of free-machining steel S12L14 is remarkably good mainly due to the fact
that the addition of Lead (Pb) to steel largely reduces the shear strength in the workpiece
and subsequently results in tighter chip curling. At lower feeds (0.1 mm/rev), the effect
of free-machining steel is even more significant in comparison with the other two
materials. Between materials CS1020 and EN36A, it can be concluded that the chip
breakability is lower if the work material contains Chromium (Cr) and Nickel (Ni) due to
the increased chip ultimate strain.
27
1.0
0.9
0.8
0.7-
0.6-
0.5
0.4-
0.3-
0.2
0.1 -
0.0
x
0 3 CO
>
O XI
E 0
Material C31Q2Q n f=o.1 mm/rev o f=o.2mm/rev
Material 512L14 • f=0.1 mm/rev • f=0.2mm/rev
Material EN36A x f=0.1mrn/rev + f=0.2mm/rev
— i 1 1 1 1 1 1 1 1 1 —
0.0 0.5 1.0 1.5 2.0 2.5 Depth of Cut (mm)
3.0 3.5
Figure 2.9 The effect of low-carbon-based alloying on chip breakability (Tool Insert: T N M G 1 6 0 4 0 8 E N A , Cutting Speed = 200 m/min)
2.4.2-4 Comparative analysis of six different work materials
4.0
Figure 2.10 shows the effect of feed on chip breakability for six work materials at
varying feeds with a typical cutting speed. In all cases except material S12L14, there
exists a sensitive feed range between 0.2 to 0.3 mm/rev, during which a small change
in feed has a significant effect on chip breakability. When the feed is higher than 0.3
mm/rev, there is no significant effect on chip breakability with the increase in feed.
Furthermore, a representative set of chip breakability diagrams in terms of standard
feed-depth of cut relationship is shown in Figure 2.11 for six work materials with
membership values p(x) = 0.5 as the boundary lines.
28
i.u -
0.9-x-"£ 0.8-
o 0.7-3 « 0.6-
o. 0.5-
"5 0.4-0 2 0.3-E ® 0.2-
0.1-
0.0-
Jfg^2+— • • — *
^iT^ Miliar
V^f Cutting Speed (V) : 100 m/min 4++ Depth of Cut (d) : 1.5 m m
Tool Insert : TNMG160408ENZ , • i - - i 1 1 — , • — i
o •
X •
+
a
= 8 6 •
CS1020
K1040
S12L14
AISI4140
EN25
EN36A
—' H 0.00 0.20 0.40
Feed 0.60
(mm/rev) 0.80 1.00
Figure 2.10 The effect of different work materials on chip breakability
• " "
Jl(X> :0.E
|i(x) 0.5
• I • 0 0.1 0.2 0.3 0.4 0.5 0.60 0.1 0.2 0.3 0.4 0.5 0.60 0.1 0.2 0.3 0.4 0.5 0.6
(1)--CS1020 (2)--K1040 (3)--S12L14
0 0.1 0.2 0.3 0.4 0.5 0.60 0.1 0.2 0.3 0.4 0.5 0.60 0.1 0.2 0.3 0.4 0.5 0.6
(4).-AISI4140 (S)-EN2S (6)-EN36A
Feed (mm/rev)
Figure 2.11 Chip breakability diagrams for six different work materials (Tool Insert: TNMG160408, Cutting Speed = 200 m/min)
29
2.4.3 Effect of Tool Inserts
Different chip breakers have different regions of chip breakability. Figure 2.12 shows
the boundary lines with membership values p(x) = 0.5 for the seven tool inserts in the
feed-depth of cut relationship diagram at higher cutting speeds (200 m/rnin). With the
same method, Figure 2.13 shows the diagrams for six tool inserts at lower cutting
speeds (50 m/min). It is noticed that there are two boundary lines at the low cutting
speed, mainly due to the complex B U E formation.
4.0-
^ 3.0 E E 3
O
2.0-
Q. 0 Q
1.0-
0.0
u.(x) < 0.5
— I • 1 1 1 —
0.20 0.30 0.40 Feed (mm/rev)
TNMG160408ENZ
TNMG160408ENA
TNMG160408ENT
TNMG160408ENG
TNMG160408ENK
TNMG160408ENJ
TNMG160408EFJ
0.00 0.10 0.50 r
0.60
Figure 2.12 Chip breakability diagram for different tool inserts at a high cutting speed (Work Material: K1040, Cutting Speed = 200 m/min)
30
4.0
3.0 -
2.0 -
1.0
0.0
- H(x)\ > \ 0.5 I
r '"• i—•—
> u(x) > 0.5
ji(x) < 0.5
- I — i — | — i 1 — i — i — i —
H(x) <
0.5
u ]i(x) > 0.5
* |i(x) < 0.5
— I — ' — l — • — i — • — i — • —
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6 (1)--TNMG160408ENZ (2)-TNMG160408ENA
4.0 F
E E
3
o
3.0
2.0
£ 1.0 a. 0 Q
H(x) < 0.5 \
u.(x) > 0.5
> H-(x) > 0.5
^(x) > 0.5
ji(x) < 0.5
0.0 | • '| • 1 • 1 1 1 • 1 1 1| • ' I r — | 1 1 r — ! 1 r
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
(3)-TNMG160408ENT (4)-TNMG 160408 EN K
4.U
3.0
2.0
1.0
00
f"tt(x) <
0.$/
• I
H(x) > 0.5
H(x) < 0.5 -1 1 r —•/— 1- "-r r- 1- -f
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
(5)-TNMG160408EFJ (6)-TNMG160408ENJ
Feed (mm) Feed (mm)
Figure 2.13 Chip breakability diagrams for different tool inserts at a low cutting speed (Work Material: K1040, Cutting Speed = 50 m/min)
J
31
A s discussed before, the chip breakability decreases with the increase of cutting speed,
however, the extent to which it is affected by the cutting speed varies with different
chip breakers. A s shown in Figure 2.14, E N A , E N G and EFJ types of tool inserts
appear to be less affected by the increase of cutting speed, giving a m u c h better chip
breakability than the other three types.
1.0
X
0 3
0 XI
E 0
0.9-
0.8-
0.7-
0.6-
0.5--
0.4-
0.3-
0.2-
0.1 -
0.0--
o ENZtype + ENG type B ENT type x EFJ type • ENA type • flat-faced
Work Material : K1040 Depth of Cut (d) : 2m m Feed (f) : 0.2 mm/rev
5 0 100 Cutting
150 Speed
-i 1 —
200 (m/min)
250 300
Figure 2.14 Comparison of cutting speed effect with different chip breakers
2.4.4 Effect of Tool Geometries
Variations in cutting tool geometry such as the rake angle, clearance angle and cutting
edge inclination angle are found to be less c o m m o n against the variations in the cutting
edge angle (Cs) and tool nose radius (r) for a given cutting tool of defined
specifications and thus were not considered in the initial analysis.
2.4.4-1 Analysis of cutting edge angle
A s shown in Figure 2.15, the cutting edge angle, defined as the angle between the
feed direction and the straight part of major cutting edge, has a significant influence on
32
chip breakability. At a high cutting speed (200 m/min), chip breakability is almost the
same in the low feed region for all cutting edge angles, while in the high feed region,
higher cutting edge angles in general produce better chip breakability. W h e n both
cutting speed and feed are very low, a rather complex effect occurs which is much the
same as what w e described before about the effect of feed at a low cutting speed (50
m/min), i.e. the likely formation of B U E . Overall, the decrease in cutting edge angle
reduces the effectiveness of chip breakers and thus lowers the chip breakability.
X,
e 3 CO
>
a 0 k.
0 XI
E 0
S
1.0-
0.9'
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0 0
(Tool
V=50m/min
o Cs = 90deg • Cs = 75 deg x Cs = 60deg •I- Cs = 45 deg -t i 1 * — — i 1 " i- i i 1 — r
.00 0.20 0.40 0.60 0.80 Feed (mm/rev)
i.u -
0.9-
0.8-
0.7-
0.6-
0.5-
0.4-
0.3-
0.2 J
0.1 -
0.0 -i
V: =200m/min
0 •
X
Cs = 90deg
Cs = 75 deg
Cs = 60 deg
Cs = 45 deg i • I '
0.00 0.20 0.40 0.60 Feed (mm/rev)
0.80
Figure 2.15 The effect of cutting edge angle on chip breakability Insert: T N M G 1 6 0 4 0 8 E N Z , W o r k Material: K1040, Depth of Cut: 2 m m )
2.4.4-2 Analysis of tool nose radius
The relationship between tool nose radius and chip breakability is shown in Figure
2.16. Generally speaking, the smaller the tool nose radius, the better the chip
breakability, while at the low cutting speed (50 m/min), larger tool nose radius
appears to have better chip breakability within the low feed range (f < O.lmm/rev).
33
The effect of tool nose radius tends to be less significant with the increase in cutting
speed.
1.0-
0.9-
E 0.8-3. 0 0.7-3 0 0.6-
a.0.5-
0 0.4-o •o 0.3-
| 0.2 J
0.1 -
v=
0.00
=50m/min
• i i - r~ T"
o r = 0.4 mm
• r = 0.8 mm x r = 1.2mm
-f r = 1.6 mm • • i
0.10 0.20 0.30 0.40 0.50
Feed (mr n/rev)
1.0 • V=200m/min
0.0' T
X
+
r = 0.4 mm r = 0.8 mm r = 1.2 mm r = 1.6 mm — i — « — r
0.00 0.10 0.20 0.30 0.40 0.50 Feed (mm/rev)
Figure 2.16 The effect of tool nose radius on chip breakability (Tool Insert: ENT-type, Work Material: K1040, Depth of Cut: 2 m m )
2.5 FOUNDATION OF A FUZZY-SET MATHEMATICAL MODEL
Factors affecting chip breakability have been studied in detail by the extensive
machining experiments. However, a quantitative utilisation of these results is required
in order to extend the prediction range of chip breakability to arbitrary combinations of
machining parameters. Therefore a fuzzy-set mathematical model is introduced
because of its ability to quantify the usually adopted descriptive words, such as slight
increase, large decrease, etc., for the effect of each machining parameter on chip
breakability. The detailed procedure for establishing such a fuzzy-set model is
described below.
34
The membership value Ufc( i, j ) will vary when the combinations of five machining
parameters (see Table 2.2) are different from those in the standard chip breakability
matrices. Knowledge rules are used to deduce the chip breakability for any
combinations of machining parameters on the basis of the established standard chip
breakability matrices. For each machining parameter, there are corresponding groups
of knowledge rules developed by using fuzzy logics. The following is an example for
such a rule used in the present work for feed:
IF Tool insert is ENT type and
Material is K1040orEN25 and
Cutting speed is > 40 m/min and
Depth of Cut is > 0.8 m m and
while Feed varies between 0.23 and 0.25 mm/rev
THEN the chip breakability, i.e. the membership value p(x) will
have a large increase comparing with that at the standard feed
value f = 0.2 mm/rev in the standard chip breakability matrix.
With the quantitative representation of chip breakability by giving membership value
p(x) for each chip shape/size, it becomes feasible to structure and describe the
machining conditions which differ from those involved in the basic chip database, and
to formulate these varying conditions in models. Fuzzy exponential function, as
shown in Figure 2.17, is used to quantify the variation of the membership value
corresponding to each different machining parameter,
u*(x) = [ p(x) ]<* (2.5)
35
where the exponent q is decided by the fuzzy linguistic values shown in Table 2.4, •ft
and p (x) is the modified membership value indicating the variation of chip
breakability.
0 3
0 XI
E 0
•o
0
•D
O
Figure 2.17 Fuzzy exponential functions
Table 2.4 Fuzzy linguistic values
1.0
A A
M
•
0
«
• +
•
o
•
q=0.5
q=0.65
q=0.72
q=0.8
q=0.9
q=1.0
q-1.1
q=1.2
q=1.35
q=1.5
q=2.0
Chip Breakability (p*(x))
very rapid increase
rapid increase
large increase
moderate increase
slight increase
no change
slight decrease
moderate decrease
large decrease
rapid decrease
very rapid decrease
Exponent q
0.50
0.65
0.72
0.80
0.90
1.0
1.10
1.20
1.35
1.50
2.0
36
For the assessment of chip breakability involving more changeable machining
parameters, membership values, u.(i, j ), can be modified by the integrated effects of
these parameters. The modified membership value, p (i, j), can be determined by the
following fuzzy-set mathematical model,
nqn
i = 1, - , 6; j = 1, «• ,6; and k = 1, - , 45;
where ql5 q2, q3, q4 and q5 are the exponents corresponding to the tool insert, work
material, cutting speed, depth of cut and feed respectively, and are determined from
the knowledge-rules that summarise all the effects of each machining parameter. O n
the basis of this fuzzy-set mathematical model, a predicting output for chip
breakability is derived with fuzzy ratings. Table 2.5 shows the likely outputs for chip
breakability as well as the corresponding chip shapes/sizes.
Table 2.5 Predicting chip breakability with fuzzy ratings
Output Ratings
1
2
3
4
5
6
7 8
Mk'&J)
0.0-0.2
0.2-0.3
0.3-0.45
0.45-0.5
0.5-0.58
0.58-0.7
0.7-0.9 0.9-1.0
Fuzzy Definition of Chip Breakability absolutely unbroken
very difficult to break
usually difficult to break
maybe broken (about 4 0 % probability)
maybe broken (about 6 0 % probability)
usually easy to break
very easy to break
always broken
The Most Likely Chip Shapes/ Sizes Produced in Machining
large snarled continue and long size with large diameter continue long size snarled with medium or large size long size (continue or broken) snarled often with few turns or small size medium size spiral with few turns short to medium size flat spiral with medium size conical spiral with medium size short size, full turn flat or conical spirals with short size side-curl arcs or up-curl arcs up-curl arcs or connected side-curl arcs
37
2.6 ESTABLISHMENT OF AN EXPERT SYSTEM FOR PREDICTING CHIP BREAKABILITY
Expert systems are computer programs that can help advise, diagnose, analyse,
consult and categorise. They simulate the problem-solving process of a human expert
on the basis of rules obtained by developing the knowledge of experts and the
database which collects facts or the relationships of facts. The programming language
used in developing the expert system is Turbo Prolog.
2.6.1 Knowledge Representation
2.6.1-1 Basic facts about chip breakability
Since the knowledge is expressed as facts and rules in Prolog, the basic chip database
is composed of 1620 facts about chip breakability. The data structure is described by
the following Prolog code:
database:
chip (membership_value, std_tool, std_material, std_speed, std_depth, std_feed)
where std_tool = standard tool inserts : ENZ type, ENA type, ENT type
std_material = standard work materials : CS 1020, K1040, EN25
std_speed = standard cutting speeds : 50,75,100,150, 200 m/min
std_depth = standard depths of cut: 0.25,0.5,1.0, 2.0, 3.0,4.0 m m
std_feed = standard feeds : 0.06, 0.1, 0.2, 0.3, 0.4, 0.5 mm/rev
38
With the above membership_value considered as elements, depth of cut as row
variable, feed as column variable, 45 standard chip breakability matrices are set up
under each combination of the above tool inserts, work materials and cutting speeds
(see Equation 2.4).
2.6.1-2 Knowledge rules for the effects of machining parameters
Based upon the standard chip breakability matrices and the analysis of results from
the experiments for investigating factors affecting chip breakability, knowledge rules
are summarised for each machining parameter. These rules are used to infer one fact
from the given facts in the basic chip database. The inference engine is based on the
established fuzzy-set mathematical model, i.e. Equation 2.6 from which a modified
membership value Pk*(i»j) is obtained. A knowledge rule takes the following general
form:
P if (Pi and P2 and P3 and and Pn) (2.7)
where P is the head of a rule which is the hypothesis and represents the goal to be
achieved, (Pi and P2 and P3 and and Pn) is the body of a rule which is a
subgoal consisting of several conditions. The head of the rule for each machining
parameter is given as follows:
tool(tool, material, speed, depth, feed, std_tool, tool_exponent)
material(std_tool, material, speed, depth, feed, std_material, material_exponent)
speed(std_tool, std_material, speed, depth, feed, std_speed, speed_exponent)
depth(std_tool, std_material, speed, depth, feed, std_depth, depth_exponent)
feed(std_tool, std_material, speed, depth, feed, std_feed, feed_exponent)
tool_Cs(std_tool, cs_angle, speed, depth, feed, cs_factor)
39
tool_NR(nose_radius, speed, feed, nr_factor)
where std_tool, std_material, std_speed, std_depth and std_feed are the standard
values in the basic chip database, and tool_exponent, material_exponent,
speed_exponent, depth_exponent and feed_exponent correspond to qlf c^, <&, q4 and
q5 in Equation 2.6 respectively. The symbols cs_factor and nr_factor are the factors
to modify the tool_exponent.
2.6.2 Examples of Knowledge Rules for Chip Breakability
Two examples of the knowledge rules for each machining parameter are given below:
(1) Rules for Tool Inserts
Example 1
In Prolog
tool(T,M,V,D,F,ena,0.8):-
T=eng,
M=M,
V<60,
D=D,
F<0.15.
-— Read As -
IF Tool insert is E N G and
Work material is any and
Cutting speed is V < 60 m/min and
Depth of cut is D = any
Feed
and
is F < 0.15 mm/rev
T H E N membership value p(x) has a moderate
increase, i.e. exponent qi = 0.8, comparing
with that at the standard tool insert ENA.
Example 2
In Prolog
tool_Cs(StdT,Tcs, V,D,F, 1.5):-
StdT=enz,
Read As
IF Standard tool insert isENZtype
Cutting edge angle Cs = 45° and
40
Tcs=45,
V<=75,
D>1.0,
F<0.25,
Cutting speed is V<75m/min and
Depth of cut is D > 1 . 0 m m and
Feed is F < 0.25mm/rev
T H E N membership value p(x) has a rapid decrease,
i.e. cutting edge angle factor = 1.5,
comparing with that at the standard
cutting edge angle Cs = 90 .
(2) Rules for Work Materials
Example 1
In Prolog
material(StdT,M,V,D,F,en25,0.9):- IF Standard tool insert is E N Z
StdT=enz,
M=en36a,
V>=60,
D=D,
F>=0.2,
F<=0.3.
Read As
and
Work material is EN36A and
Cutting speed is V > 60 m/min and
Depth of cut is D = any and
Feed is 0.2 < F < 0.3 mm/rev
T H E N membership value p(x) has a slight
increase, i.e. exponent q2 = 0.9,
comparing with that at the standard
work material EN25.
Example 2
In Prolog
material(StdT,M,V,D,F,csl020,0.65):
StdT=StdT,
M=sl2114,
V>=60,
Read As
IF Standard tool insert is any and
Work material is S12L14 and
Cutting speed is V > 60rn/min and
Depth of cut is D > 2 . 5 m m and
41
D>2.5,
F<0.2.
Feed is F < 0.2mm/rev
T H E N membership value p(x) has a rapid
increase, i.e. exponent qi = 0.65,
comparing with that at the standard
work material CS1020.
(3) Rules for Cutting Speeds
Example 1
In Prolog
depth(StdT,StdM,V,D,F, 150,0.9):-
StdT=enz,
StdM=kl040,
V>=130,
V<145,
D>2.0,
F>=0.2,
F<=0.23.
Read As
IF Standard tool insert isENZ and
Standard work material is K1040 and
Cutting speed is
130 £ V < 145 m/min and
Depth of cut is D > 2.0 m m and
Feed is 0.2 < F < 0.23 mm/rev
T H E N membership value p(x) has a slight
increase, i.e. exponent q3 = 0.9,
comparing with that at the standard
cutting speed V = 150 m/min.
Example 2
In Prolog
depth(StdT,StdM,V,D,F,50,l.l):
StdT=StdT,
StdM=StdM,
V>=52,
V<60,
— Read As
IF Standard tool insert is any and
Standard work material is EN25 and
Cutting speed is
52 < V < 60 m/min and
Depth of cut is D = any and
Feed is F < 0.1 mm/rev
42
D=D,
F<0.1.
T H E N membership value p(x) has a slight
decrease (i.e., exponent q3 = 1.1)
comparing with that at the standard
cutting speed V=50 m/min.
(4) Rules of Depths of Cut
Example 1
In Prolog Read As
depth(StdT,StdM,V,D,F,2.0,1.35):- IF Standard tool insert is E N Z
StdT=enz,
StdM=StdM,
v=v,
D>1.6,
D<=1.8,
F>=0.25.
and
Standard work material is any and
Cutting speed is V = any and
Depth of cut is 1.6 < D < 1.8 m m and
Feed is F > 0.25 mm/rev
T H E N membership value p(x) has a large
decrease (i.e., exponent q4= 1.35)
comparing with that at the standard
depth of cut D = 2.0 m m .
Example 2
In Prolog
depth(StdT,StdM,V,D,F, 1.0,0.8):
StdT=ent,
StdM=csl020,
V=V,
D>1.5,
D<2.0,
F>=0.2,
IF
Read As
T H E N
Standard tool insert is E N T and
Standard work material is CS1020 and
Cutting speed is V=any and
Depth of Cut is 1.5 < D < 2.0 m m and
Feed is 0.2 < F < 0.25 mm/rev
membership value p(x) has a moderate
increase, i.e. exponent q4 = 0.8,
43
F<0.25. comparing with that at the standard
depth of cut D = 1.0 m m .
(5) Rules for Feeds
Example 1
In Prolog
feed(StdT,StdM,V,D,F,0.2,0.8,l):-
StdT=ena,
StdM=en25,
V>=40,
V<60,
D>0.8,
F>0.2,
F<0.25.
Read As
IF
T H E N
Standard tool insert is E N A and
Standard work material is EN25 and
Cutting speed is
40 < V < 60 m/min and
Depth of cut is D > 0.8 m m and
Feed is 0.2 < F < 0.25 mm/rev
membership value p(x) has a moderate
increase, i.e. exponent qs = 0.8,
comparing with that at the standard
feed F = 0.2 mm/rev.
Example 2
In Prolog Read A s - —
feed(StdT,StdM,V,D,F,0.1,1.0,l):- IF Standard tool insert isENZ
StdT=enz,
and
StdM=StdM,
V>=40,
D>2.0,
F>0.1,
F<=0.18.
Standard work material is any and
Cutting speed is V £ 40 m/min and
Depth of cut is D > 2.0 m m and
Feed is 0.1 < F < 0.18 mm/rev
T H E N membership value p(x) has no change
, i.e. exponent qs = 1.0, comparing
with that at the standard feed F=0.1
mm/rev.
4 4
2.6.3 Schematic Diagram of the Expert System
On the basis of what was discussed before, a schematic diagram of the expert system
can be drawn as shown in Figure 2.18.
Basic Machining Experiments
I Basic Chip Database
Fuzzy Membership
Function
Standard Chip
Breakability Matrices
INPUT
(1) Tool Insert Type (2) Work Material (3) Cutting Speed (4) Depth of Cut (5) Feed (6) Cutting Edge Angle (7) Tool Nose Radius
<.
Extensive Machining Experiments on the Effect of Each Machining Parameter
I Knowledge Rule Base for (1) Effect of Each Machining
Parameter on Chip Breakability
(2) Chip Shapes/Sizes
o Inference Engine
Fuzzy-Set Mathematical Model
I OUTPUT
(1) Chip Breakability
(2) Likely Chip Shapes/Sizes
USER >
Figure 2.18 The expert system schematic diagram
45
2.7 PROLOG PROGRAMMING ARCHITECTURE
Prolog stands for "PROgraming in LOGic" and is a computer programming language
which was invented in 1970s. Prolog is powerful in handling logic problems and
particularly well suited for applications in the artificial intelligence systems, such as,
expert systems. Turbo Prolog is one of Prolog versions used on an I B M P C
computer (or compatible) system.
Considering the large scope of the expert system program, a modular architecture is
introduced in programming, i.e. the whole program is divided into several modules
according to their functions. Then a global method is used to communicate across
module boundaries. All the data items are stored in the external database in the form
of chains which allow the expert system to access the data items very quickly. Figure
2.19 shows the programming structure of the expert system.
In this way, the program structure has the following advantages:
(a) With the basic chip database as external database in files, the only limit on the
database size is the available computer disk space which is usually large enough
to hold any practical database.
(b) With all the data stored as chains in the external database, the expert system is
able to handle efficiently even a very large amount of data on disk.
(c) With modular programming, the expert system is very easy to develop new
functions only by creating new modules without influencing on other modules.
46
r RULES for Chip Breaker
RULES for Work Material
RULES for Likely Chip Shapes/Sizes
GLOBAL
Program
f RULES for Cutting Conditions
RULES for Tool Geometry
Main
Program
External
Database
Figure 2.19 P R O L O G program structure diagram
2.8 SUMMARY AND CONCLUDING REMARKS
(1). Advances in automated machining systems inevitably require more effective
chip control. Particularly, predictability of chip breaking has been identified as
vital since the performance of chip control greatly depends on the chip
breakability in machining of steel.
(2). Chip breaking is a complicated process, influenced by a great number of
interacting process variables. The theory and knowledge to date about metal
cutting and chip control are not available to predict quantitatively the chip
breakability for the application to workshop environments.
(3) By introducing a fuzzy membership function, this chapter has presented a new
method to quantify chip breakability according to the chip shapes/sizes
produced in the machining process. Based on the basic chip database for
47
providing the standard chip breakability and the extensive machining
experiments for evaluating the effect of each machining parameter, an expert
system has been developed using a fuzzy-set mathematical model for
quantitatively predicting the chip breakability under any combinations of cutting
conditions, work materials and tool design features including tool geometry.
(4). The prediction accuracy of chip breakability depends heavily on the range of
experimental results available in the "knowledge-base" and on the quality of
knowledge rules that could be developed.
(5). The method for predicting chip breakability presented in this chapter may
provide a quantitative reference for optimal machining process control or
machinability assessment, and also will provide a feasible application to total
chip control in unmanned machining systems in the future.
48
CHAPTER 3
A PREDICTIVE EXPERT SYSTEM FOR THE ASSESSMENT OF MACHINING PERFORMANCE WITH CHIP CONTROL AS A MAJOR CRITERION
3.1 INTRODUCTION
The development of automated process planning (APP) is of great interest to
C A D / C A M integration. Developing expert systems for machining parameter selection
is an effective approach for improving A P P [64-66]. However, chip control is seldom
selected as a criterion for A P P mainly due to its complexity. Aiming at providing an
effective auxiliary to A P P system, this chapter presents the development of a predictive
expert system for the assessment of machining performance with chip control,
including chip breakability/chip forming patterns, as a major criterion and with due
consideration of surface finish and power consumption.
The predictive expert system is set up based on the results of a series of experimental
work, the knowledge rules derived from the present knowledge on chip control and
machining, and the expert's knowledge and findings. The knowledge-base created
includes a comprehensive database containing basic facts about chip breakability,
surface finish and power consumption, and a set of knowledge rules covering the
effects of different tool inserts, work materials, cutting conditions and tool geometries.
49
3.2 FOUNDATION OF MACHINING PERFORMANCE DATABASE
A database is an essential part of the predictive expert system and has the following
functions:
(a) storing basic facts about machining performance based on machining
experimental results;
(b) being used as a primary standard for assessing machining performance;
(c) being used as an analytical criterion for the knowledge rule summary; and
(d) providing referable information during the expert's consultation process.
A series of machining experiments on chip control, surface finish and power
consumption was conducted to set up the machining performance database for a wide
range of cutting tools, work materials, cutting conditions and major tool geometries,
shown as follows.
(1) Cutting Tools :
(i) one conventional flat-faced tool (without chip breaker);
(ii) eight commercial tool inserts with different types of chip breakers (ENA,
E N Z , ENT, ENK, E N G , ENJ, EFJ and EFB types ).
(2) Work Materials :
(i) low carbon steel - CS 1020 (BHN=185)
(ii) medium carbon steel - K1040 (BHN=198)
(hi) high carbon steel - K1055 (BHN=220)
(iv) free-machining steel - S12L14 (BHN=168)
(v) Ni-Cr alloy steel - EN36A (BHN=210)
(vi) Mo-Cr alloy steel - AISI4140 (BHN=300)
(vii) Ni-Mo-Cr alloy steel-EN2i (BHN=240)
50
(3) Cutting Conditions :
(i) cutting speed = 100 ~ 350 m/min
(ii) depth of cut = 0.25 - 5 m m
(iii) feed = 0.06 - 1.0 mm/rev
(4) Tool Geometries :
(i) tool nose radius = 0.4, 0.8, 1.2 & 1.6 m m
(ii) cutting edge angle = 90,75, 60 & 45 degrees
(iii) tool rake angle = -5 & + 5 degrees
(iv) tool inclination angle = -12, -6,0 & +5 degrees
3.3 ASSESSMENT OF MACHINING PERFORMANCE
Chip control, including chip breakability/chip forming patterns, combined with the
power consumption and surface finish will give a better assessment for machining
performance.
3.3.1 Assessment of Chip Control
The major objective of chip breaking is to facilitate effective and hazard-free chip
disposal. In fact, the chip control operation requires efficient breaking of the chips
into acceptable sizes and shapes for disposal in automated machining systems.
Introducing a factor for chip acceptability therefore becomes vital from the point of
view of chip disposal and safe machining. Chip acceptability can also be related to the
chip shapes/sizes produced. Let us define the chip acceptability factor g(x) as
follows:
5 1
g(x) = k, k=l,...,7 (3.1)
where x is the chip shape (the same as that shown in Equation 2.2 in Chapter 2) and
k is a positive integer representing the different grade of chip acceptability. The larger
the chip acceptability factor g(x), the more acceptable is the chip for chip disposal. A
range of g(x) is shown in Table 3.1.
Table 3.1 Chip acceptability factor g(x) in automated machining systems
Chip Shapes/sizes
large snarled chips
continuous chips
tooth-edged chips
long size size chips
broken medium snarled chips
broken long size chips
medium size chips
small snarled chips
short size broken chips
small arc chips
*?(*) 1
2
3
4
5
6
7
Definition and Explanation
most unacceptable due to the main
reason for inducing tangled chips
unacceptable even if the chips are
broken chips due to high cutting
force, poor surface finish, etc.
usually unacceptable
maybe acceptable depending on the
practical machinina requirements
acceptable although chips are not
broken due to easiness of chip disposal
usually acceptable
most acceptable
Chip acceptability factor g(x) can be deduced from the knowledge rules for chip
shapes. Table 3.2 shows the integrated (combined) assessments for chip control after
considering both the chip breakability and the chip acceptability. The most likely chip
shapes/sizes are also provided to complement the basic understanding.
52
Table 3.2 Integrated assessment for chip control in automated machining systems
No.
1
2
3
4
5
6
7
8
9
10
1 1
12
Mx)
0.0--0.2
0.2--0.3
0.3--0.45
0.45--0.5
0.5--0.58
0.58--0.7
0.7--0.9
0.9-1.0
0.0--0.5
0.5 -0.85
0.85-1.0
0.3--0.45
g(x)
1
1
3
4
4
6
7
7
2
2
2
5
Integrated Outputs
absolutely unbroken
and most unacceptable
very difficult to break
and unacceptable
usually difficult to break
and usually unacceptable
maybe broken (about
40% probability)
and maybe acceptable
maybe broken (about
60% probability)
and probably acceptable
usually easy to break
and usually acceptable
very easy to break
and most acceptable
always broken and often
overbroken; though most
acceptable for disposal,
usually not suggested
unbroken
and unacceptable
broken
but unacceptable
overbroken and
absolutely unacceptable
due to very harmful to
the machining system
unbroken
but acceptable
Most Likely Chip Shapes/sizes
large snarled chip
infinite chip
tubular chip with large diameter
continuous long size chip
medium-large size snarled chip
long size (continuous or broken)
snarled (often with few turns)
medium size chip
spiral chip with few turns
short to medium size chip
medium size flat spiral chip
medium size conical spiral chip
short size chip
full turn chip
short size flat spiral chip
short size conical spiral chip
side-curl arcs
up-curl arcs
very tight up-curl arcs
connected side-curl arcs
long size tooth-edged chip
short to medium size and
tooth-edged chip
connected side-curl arc type
with heavily tooth-edged chip
small snarled chip
53
3.3.2 Assessment of Surface Finish
The quality of machined surface is very closely associated with the chip shapes
produced. Steady and continuous chips will produce a better surface finish while
broken and tightly curled chips result in worse surface finish. In general, the
formation of built-up edge (BUE) will cause deterioration of surface finish due to the
effect of B U E fragments being randomly deposited on the machining surface.
Furthermore, poor chip control results in tangled chips which, in turn, will scratch the
machined workpiece surface and even lead to unexpected machine tool or cutting tool
damage.
The surface finish was measured by Surfcom Model 550AD Surface Measuring
Equipment, with the arithmetic mean deviation, Ra, as an assessment criterion. Four
Ra values on the periphery, but along the same circle of workpiece, were taken to
obtain an average Ra value. Some representative data of surface finish for different
tool chip breakers are given in Appendix B.
3.3.3 Assessment of Power Consumption
The power consumption has a direct effect on the heat production, further influencing
tool wear, tool life and the machined surface quality. It has been shown that the
effective chip breaking at reduced power consumption can be achieved by means of
efficient tool groove geometries [60].
Power consumption rate was assessed by evaluating the main cutting force measured
by the KISTLER Dynamometer (Type 9257A).
54
3.4. AN INVESTIGATION OF THE INFLUENCING FACTORS ON MACHINING PERFORMANCE
A s the factors influencing chip breakability have been investigated in detail in Chapter
2, in this section, the analysis will be emphasised on surface finish and power
consumption.
3.4.1 Effect of Tool Insert Types
The performance of each tool insert was tested through machining experiments for
chip breakability, power consumption and surface finish. Different chip breakers give
different chip forms for a given set of cutting conditions, while producing varying
surface finish. Figure 3.1 shows a representative result reflecting this.
2.5
E
>
o dd.
E
2.0-
n EFB-type x EFJ-type o ENJ-type
+ ENK-type
0.00
Work Material : CS1020 U=175m/min d=0.5mm
0.10 0.15 Feed (mm/rev)
Figure 3.1 Surface finish comparison for four tool inserts
A s shown in Figure 3.1, the groove-type chip breaker with the smaller restricted
contact length (EFJ, h = 0.1mm) has a better surface finish than that of the larger
55
restricted contact length (ENJ, h = 0.25mm). E F B type has the best surface finish
mainly due to its special configuration of chip breaker.
Reducing power consumption is of great importance for achieving improved tool life
and efficient production. The chip breaker configuration and the restricted contact
length are the two major factors to be considered in the design of efficient chip
breakers. Shown in Figure 3.2 is the power comparison made among several
representative tools.
800 I • i 1 1 • i ' i • 1 • r—' 0.00 0.10 0.20 0.30 0.40 0.50 0.60
Feed (mm/rev)
Figure 3.2 Power consumption comparison for five tool inserts
It is apparent from Figure 3.2 that the flat-faced tool consumes the highest power
while the groove-type tool with smaller restricted contact length (EFJ) consumes the
least power. Between the two representative chip breakers, i.e., the groove-type
(ENT) and the obstruction-type (ENG), E N T proves to be better for low feed rates
while E N G is better for high feed rates.
56
3.4.2 Effect of Work Materials
The effect of work material properties including chemical composition and material
hardness on machining performances is very complex to investigate due to the
difficulties in isolating the effects of various alloying elements in the work specimens.
In this section, six typical work materials were selected for analysis. A s given below
in Figure 3.3 and 3.4 are two representative experimental results of surface finish and
power consumption for different work materials.
3.4.3 Effects of Cutting Conditions said Tool Geometries
O n the basis of the machining experiments and the knowledge derived from the
published work, an integrated summary for the effects of cutting conditions and major
tool geometries on machining performance is given in Figure 3.5.
< 0.8-4 1 1 j 1 1 1 , 1 1 1 4 1 1—
0.06 0.08 0.1 0.12 0.15 0.2 0.25 Feed (mm/rev)
Figure 3.3 Surface finish comparison for four work materials
1400
o "• 1200-u
1000-
3
° 800-
600
Tool Insert : ENT-type V=100m/min d=2.0mm
o CS1020
+ K1040
• EN36A
X AISI4140 — I • 1 ' I • 1 •» 1 » 1 » i —
0.06 0.1 0.2 0.3 0.4 0.5 0.6 Feed (mm/rev)
Figure 3.4 Power consumption comparison for four work materials
Parameters
(X-axis)
Chip Breakability
H(x) (Y-axis)
Surface Finish Ra(um)
(Y-axis)
Power Consumption
Fc(N) (Y-axis)
Cutting Speed
significant significant slight
m) very significant Feed very
significant
D e p t h of Cut
significant for small d
significant for very small d
. very significant
* i significant ^ significant Nose Radius
significant
4 slight ii significant Cutting Edge Angle
significant
4 slight i i significant Rake Angle
Inclination Angle
nearly constant
slight significant
Figure 3.5 Integrated effects of cutting conditions and tool geometries
58
3.5 EVALUATION OF WORK MATERIAL PROPERTIES
The properties of work materials have significant yet complicated effects on machining
performance including chip breakability, surface finish and power consumption. In
Chapter 2, the effect of work material on chip breakability has been studied sufficiently.
In order to describe comprehensively the effects of work materials, a work material effect
table is introduced for various classes of machining conditions, aiming at providing
referable information for automated process planning systems. Shown in Table 3.3 is an
example of work material EN25 when machining with a conventional groove-type tool
insert (ENT type).
Table 3.3 A n example of work material effect table (Tool Insert: E N T type)
The Classes of
Machining Conditions
1
2
3
4
5
6
Precision Finishing
Normal Finishing
Semi-Finishing
Light Roughing
Normal Roughing
Heavy Roughing
Chip
Breakability
very difficult to break
very difficult to break
usually difficult to break
maybe or partly broken
usually easy to break
very easy to break
Surface
Finish
good
fair
fair
acceptable
poor
very poor
Power
Consumption
low
moderate
moderate
moderate
high
very high
Furthermore, the effects of work materials on surface finish and power consumption can
be quantitatively evaluated by introducing a fuzzy membership function. Table 3.4
shows the fuzzy ratings of surface finish and power consumption based on the
conventional machining requirements.
59
Table 3.4 Fuzzy ratings of surface finish and power consumption
Assessment of
Surface Finish
Range
Ra(pm)
< 0.6
0.6 - 1.1
1.1 - 1.5
1.5 - 2.0
2.0 - 3.0
>3.0
Fuzzy
Description
excellent
good
fair
acceptable
poor
very poor
Membership
Value Ps
0.85 - 1.0
0.7 - 0.85
0.6 - 0.7
0.5 - 0.6
0.25 - 0.5
0 - 0.25
Assessment of
Power Consumption*
Range
^cutting (N)
<300
300-700
700 -1200
1200 -1500
1500 - 2000
>2000
Fuzzy
Description
very low
low
moderate
high
very high
unacceptable
Membership
Value pp
0.9 - 1.0
0.8-0.9
0.65 - 0.8
0.50 - 0.65
0.3 - 0.5
0 - 0.3
* The ratings of power consumption may change with different machine-tool structures.
In this way, the work material effect table can be formulated into a work material
evaluation matrix, providing an approach to assessing the deviation of membership value
due to the change of conditions. The work material evaluation matrix, W L , is
constructed with its column representing different classes of machining conditions (Table
3.3), and its row representing the fuzzy ratings of chip breakability (p(x)), surface finish
(Ps) and power consumption (pp) respectively, corresponding to a specific combination
of one work material and one tool insert with the standard tool geometry.
W T =
Pl(x) p£x) pjx) pj(x) pjx) p^x)
Us, w w
Up
H s3
w Up,
Hs< m4 W W W
Hp 4 ^ p 5 Vpt
V
w
(3.2)
L = 1, 2, ... , 63
60
The meaning of each symbol in the matrix is defined as follows :
integer L : combinations of different weak materials and the tool inserts with
standard tool geometry (0°, 5°, -6°, 90°, 60°, 0.8)
subscript number 1 to 6: six different classes of machining conditions
exponents u, v and w : exponents for deterrnining the variations of membership
values due to the change of tool geometry.
3.6 KNOWLEDGE REPRESENTATION
3.6.1 Basic Facts Based on the Database
A database was established using the results from selected machining experiments,
providing the basic facts about machining performance including chip breakability,
surface finish and power consumption. The database structure developed in P R O L O G
is shown as follows :
machinmg_performance_data (output (chip_breakabiUty_membership_value,
surface_finish, main_cutting_force), tool_insert_type,
work_material, cutting_conditions (speed, depth,
feed), tool_geometry (nose_radius, rake_angle,
inclination, angle, cutting_edge_angle) ).
Given below is an example for one of the facts established:
machining^performance.data (output(0.7,1.65,1.2), ena_type_insert,
kl040, cutting_conditions(100, 2.0, 0.2),
tool_geornetry(0.8,-5, -6, 90) ).
The fact is explained as if the tool insert type is ENA-type, work material is medium
carbon steel K1040, cutting speed =100 m/min, depth of cut =2.0 m m , feed =0.2
mm/rev, tool nose radius =0.8 m m , tool rake angle = -5°, tool inclination angle =-6°
61
and tool cutting edge angle =90°, then the chip breakability (membership value) p(x)
=0.7, surface finish R a =1.65 p m and main cutting force F c =1.2 K N .
3.6.2 Knowledge from Expert's Experience
The major objective of the expert system is to formalise the experience of human
experts and make it scientifically useable. The rule-based approach is an effective
method which uses a large storage of IF-THEN rules simulating the expert's
reasoning. Given below are some examples of expert's knowledge and experience on
machining operation and chip control.
(1) Example of expert's knowledge for the effect of feed on surface finish
It is well known that the lower the feed rate, the better the surface finish. However,
the expert's knowledge is not limited to this. From his experience on machining, more
knowledge rules can be developed as follows :
IF the cutting speed is high enough to eliminate the formation of built-up edge
T H E N the surface finish will be improved with the decrease of feed rates
BUT
IF the cutting speed is not high enough
T H E N within the low feed range, decreasing feed rate will give adverse results on
surface finish due to the marks of B U E fragments on the machined surface.
(2) Example of expert's experience for using tool inclination angles
The appropriate use of tool inclination angle is important to the machining process.
Given below are .some of the expert's strategies for using tool inclination angles:
62
(a) Use positive inclination angles for fine machining so that the chip will flow
towards the unmachined surface.
(b) Use negative inclination angles for interrupted machining to protect the tool tips.
(c) Use negative inclination angles for tempered and hardened materials to enhance the
tool strength.
(d) Use positive inclination angles when the system setup lacks rigidity and strength
to reduce the possible vibration in machining.
(3) Example of knowledge rules for answering questions
Knowledge rules are summarised to response to some specific questions, such as
"why unbroken", "why tooth-edged chips", "why poor surface finish" etc. The
following is an example of these rules (written in Prolog).
why_unbroken (Tool, Material, Speed, Depth, Feed):-
Tool = enz, Material = en25, Speed = Speed,
Depth <= 1.5, Feed >= 0.3,
write("~Because depth of cut is too small for cutting"),nl,
write("~EN25 alloy steel when using E N Z type tool."),nl,
write("~We suggest that at least d > 1.5 m m should"),nl,
write("--be used to make the chip break efficiently.").
(4) Guidance for Selecting Efficient Cutting Conditions
The optimal selection of cutting conditions is very complicated due to numerous
factors involved. The predictive expert system has the function to help users select
efficient cutting conditions in terms of machining performance. For example, in some
cases, the chip is broken but it is not an efficient machining due to too high cutting
63
force, while in the other cases, such as in finishing-machining, the chip is very
.difficult to break due to the small chip thickness. In order to resolve these sorts of
problems, some useful knowledge rules have been developed and summarised to give
a general guidance for the user to choose the most effective tool insert, tool geometry
and cutting conditions to satisfy his/her specific machining requirements. Given
below in Figure 3.6 is a sample for the .selection of efficient feed value.
USER INPUT Tool Insert, Work Material Cutting Speed, Depth of Cut
I SHOW THE IMPORTANT ORDER FOR USER TO SELECT
Chip Breakability Tool Life Power Consumption Surface Finish Dimensional Accuracy
f Determine Primary Feed Value }
-—n Check Database and Knowledge Rules to Predict the Machining Performance C D
Change the Feed Value
or Reorganise the Primary
Input Conditions if Permitted
Output the Efficient Feed Value for the Proposed Cutting Process
Figure 3.6 Flow chart for selecting efficient feed value
64
3.7 ESTABLISHMENT OF A PREDICTIVE EXPERT SYSTEM
Figure 3.7 shows a simplified schematic diagram for establishing the predictive expert
system for the assessment of machining performance. The machining performance
database created based on the machining experiments and the expert's knowledge and
experience are combined to set up the knowledge base.
Machining Experiments
I Machining Performance Database
1. Chip Breakability /Chip Forming Patterns
2. Surface Finish 3. Power Consumption
X c KNOWLEDGE BASE Z Knowledge Rules for the Effects of
1. Tool Insert Types 2. Work Materials 3. Cutting Conditions 4. Tool Geometries
I
Expert's Knowledge
and Experience
Knowledge Rules for Suggesting Improvements
I f Predictive Expert System for the Machining Performance Assessment )
Figure 3.7 A schematic diagram for setting up the predictive expert system
The established expert system provides three major functions as shown in Figure 3.8,
i.e., assessment of machining performances for user's self-selected input machining
65
parameters; suggestions for improvement on the initial input parameters according to
the user's requirement; reply to user's questions and explain the output results. In this
way, the predictive expert system provides quantitative evaluations for the automated
process planning (APP) system.
c AUTOMATED PROCESS PLANNING SYSTEM Z INPU T 1
Tool Insert Type Work Material Cutting Conditions Tool Geometry
C X
I INPUT 2
1. Objects to be improved
2. Changeable Initial Input Parameters
I
i INPUT 3 Questions concerning
the Machining Performance
Z PREDICTIVE EXPERT SYSTEM FOR THE ASSESSMENT OF MACHINING PERFORMANCE
z OUTPUT 1
Assessment of Chip Breakability, Surface Finish and Power Consumption
i OUTPUT 2
Suggestions for
Improvement
x )
OUTPUT3 Reply USER'S Specific Questions
Figure 3.8 The predictive expert system acting as an auxiliary for A P P system
3.8 SUMMARY AND CONCLUDING REMARKS
(1) A new approach for evaluating machining performance (i.e. assessing chip
breakabihty/chip forming patterns with due consideration to power consumption
rates and the surface finish produced) has been presented in this chapter. The
66
method uses a predictive expert system which was developed by using expert's
knowledge on machining operation and the results of extensive experimental
work.
(2) An interactive user-friendly program is added to this predictive system to help the
user with his/her specific queries and aimed at improving machining
performance. The basic guidelines presented in this chapter are designed to
emphasise the need for implementing this predictive expert system in an
automated process planning system.
(3) The success of this expert system depends very heavily on the applicability range
of experimental results covering all likely combinations of work materials, tool
inserts, cutting conditions and tool geometries. Therefore it is essential to have
"dedicated knowledge pools" to suit the individual requirements; and this can be
done without significant effort using the guidelines given in this chapter.
CHAPTER 4
A KNOWLEDGE-BASED SYSTEM FOR DESIGNING EFFECTIVE GROOVED CHIP BREAKERS BASED ON THREE-DIMENSIONAL CHIP FLOW IN MACHINING
4.1. I N T R O D U C T I O N
Continuing R&D efforts on cutting tools have now produced a great number of
different chip breaker configurations. However, the current practice in tool chip
breaker design and application is very heavily dependent upon the arbitrary
experimental work, which is primarily based on the historically known "try and see"
methodology. This practice is very time-consuming and does not always produce the
most desirable results in terms of the chip breakability aspect at optimum chip breaker
utilisation and the associated minimum power consumption. Therefore, a more
scientific and effective approach for designing tool chip breakers is highly called for,
against the current practice of "try and see" methods. The need for developing a
knowledge-base from a database system has been shown as essential both for
designing effective chip breakers and for achieving the total chip control in automated
machining [2].
The first attempt has been made to develop a knowledge-based approach to designing
grooved chip breakers for providing effective chip breaking at minimum power
consumption in two-dimensional chip up-curl modes in a recent work [67]. This
work has now been extended to include a more common three-dimensional chip flow,
i.e. combined chip sideflow and chip backflow (or chip streaming), for various work
materials at a much wider range of tool geometries and cutting conditions.
It has also been shown that the basic mechanism of chip flow in a majority of
commercial grooved chip breakers can be explained as resulting from the combined
effects of tool restricted contact and toolface configurations, which quite often
represent a groove with varying sizes and profiles [61]. Therefore, in the present
work, typical grooved chip breakers (i.e. groove profiles with raised back wall,
standard back wall, reduced back wall & no back wall) with various sizes and
restricted contact lengths are selected to form a reference model for a primary
systematic analysis, which could then be extended to more complicated toolface
configurations.
4.2 PRESENT KNOWLEDGE OF CHIP CONTROL
WITH GROOVED CHIP BREAKERS
4.2.1 Tool Restricted Contact Effect on Chip Breaking
The chip streaming (i.e. chip backflow) effect has been found to be playing a
significant role in chip curling and subsequent chip breaking when machining with
grooved chip breakers having a tool restricted contact adjacent to the grooved profile
[61]. For a given set of cutting conditions, the tool restricted contact length h is the
most important factor in determining the chip backflow angle T|b. Machining
experiments with restricted contact tools have shown that there exist the following
three basic models for chip streaming (backflow) [67], as shown in Figure 4.1.
Figure 4.1 Three chip streaming models
These three chip streaming models can be summarised into the following equation :
rjb < a, when h > hn
(chip curls upwards before reaching the chip groove);
% = ct, when h = hn
(chip straightening occurs); (4.1)
% > a, when h < hn
(chip enters the chip groove as a result of chip streaming).
where h is the tool restricted contact length, hn is the tool/chip natural contact length
and a is the tool rake angle.
4.2.2 Three-dimensional Chip Flow
Many researchers have presented cutting models for predicting chip flow in machining
[44-45, 68-72]. All these investigations were based on the use of flat-faced cutting
tools. However, the chip flow in practical oblique machining for the grooved tools
with varying tool nose radii, tool inclination angles and rake angles, is far more
complex than that for the flat-faced tools. Furthermore, the chip flow when
macnining with the grooved tools having various toolface configurations is even more
complicated. When dealing with a grooved chip breaker with varying tool restricted
contact lengths, the effect of three-dimensional chip flow can be presented by two
components, chip sideflow angle TJS and chip backflow (stream) angle %. As shown
in Figure 4.2, T|s is the angle formed by the chip side-curling projected onto the X-Y
plane, while t]^ is the angle formed by the chip up-curling projected onto the X-Z
plane.
Tls T~7 Tib /
^ '
\ \
Grooved Chip Breaker
\
xy
/ V x z
/
Vc
rz Figure 4.2 Three dimensional chip flow with a grooved chip breaker
4.2.3 Power Consumption Rate
The aim of studying power consumption rate is to provide design criteria for achieving
efficient chip control at reduced power consumption. Early work has emphasised the
possibility of reducing power consumption with grooved chip breakers [68, 73-76].
The tool restricted contact effect has been shown to be the most important factor for
achieving efficient chip breaking and reduced power consumption with grooved chip
breakers [60, 67, 77-78].
4.3 ESTABLISHMENT OF A DATABASE SYSTEM
The basic database established in this chapter is mainly based on the previous work
[59-60, 63,67] and provides the analytical foundation for chip breaking performance,
power consumption, knowledge rules and designing criteria for grooved chip
breakers. The database consists of three parts : Reference Database, Grooved Chip
Breaker Database and 3-D (three-dimensional) Chip Flow Database.
4.3.1 Reference Database
A series of experiments was conducted using the secondary rake type restricted
contact tools as the standard tools representing a wide range of cutting speeds (4
values), undeformed chip thicknesses (3 values), tool rake angles (5 values) and tool
restricted contact lengths (4 values). Medium carbon steel was used as the work
material in the experiments. The chip backflow angle Thj, chip thickness t2 and the
cutting forces F c and Ft measured from the experiments, were recorded to set up the
Reference Database.
4.3.2 Grooved Chip Breaker Database
Four chip groove styles (A, B, C & D) as shown in Figure 4.3, each with three
groove sizes (GT1, G T 2 & GT3), were used in the experiments on establishing the
effects of different groove profiles for various ti and h values at the cutting speed
V = 100 m/min, rake angle a = 0° and depth of cut w = 3.0 m m . The cutting forces
Fc and Ft, chip breaking performance and chip shape were recorded to set up the
Grooved Chip Breaker Database.
(a) Style A (b) Style B (c) Style C (d) Style D
Figure 4.3 Configurations of grooved chip breakers (based on [67])
4.3.3 3-D Chip Flow Database
A series of oblique cutting experiments was conducted with standard grooved chip
breakers for combinations of 3 work materials, 3 tool nose radii, 5 depths of cut and 6
feeds at a cutting speed V = 100m/min, tool inclination angle i = -6° and rake angle a
= -5°. The chip sideflow angle TJS, chip backflow angle rjb and chip thickness t2,
were measured to set up the 3-D Chip Flow Database.
A second series of experiments was conducted to establish the effects of tool
inclination angle i, cutting speed V on chip sideflow angle rjs. These results are
included in the 3-D chip flow database.
4.4 SUMMARY OF PRESENT KNOWLEDGE
FOR SETTING UP A KNOWLEDGE BASE
An attempt was made to integrate the results derived from the databases established
with the present knowledge and research findings on chip control, and then to
summarise the total knowledge into two major groups, i.e. chip breaking and power
consumption.
4.4.1 Analysis of Chip Breaking
It has been found that the chip breaking is generally influenced by the combined
effects of the chip backflow angle J]^, chip sideflow angle rjs, chip thickness t2 and
the chip shape produced.
4.4.1-1 Results of machining with the secondary rake restricted
contact tools
A series of machining experiments with secondary rake restricted contact tools was
conducted to study the individual effects of undeformed chip thickness ti, restricted
contact length h, rake angle a and cutting speed V. The knowledge acquired was
incorporated into the database for grooved chip breakers. The relationships between
these parameters and the chip backflow angle % and chip thickness t2 were plotted
for all combinations and a representative diagram is shown in Figure 4.4 to give a
general understanding of these effects.
VARIABLE
Undeformed Chip
Thickness
Restricted Contact Length
Tool Rake Angle
Cutting Speed
EFFECT ON CHIP BACKFLOW ANGLE
RELATIONSHIP
fe
fe
k Tib
a
fe
EFFECT RATING
Significant
Significant
small
small
EFFECT ON CHIP THICKNESS
RELATIONSHIP
-M2 y fe
M t 2
>H2
fe
*t2
V
w
EFFECT RATING
Significant
small
small
small
Figure 4.4 Typical relationship diagrams for restricted contact tools
4.4.1-2 Results of machining with different configurations of
grooved chip breakers
(a) Effect of restricted contact length, h
A s seen from Figure 4.5, lower h values associated with higher chip backflow
angles result in reduced chip up-curl radius, thus producing tighter and more
efficiently broken chips.
(b) Effect of groove sizes
O n the basis of the analysis of chip up-curl radius, it has been seen that the chip
breaker with smaller groove size has a better chip breakability due to the smaller chip
up-curl radius which facilitates the chip breaking. Figure 4.6 shows the variation of
chip up-curl radius with tool restricted contact length for the three chip-breaker groove
sizes tested. It also appears that the minimum chip up-curl radius ru equal to the tool
groove radius Ro will occur when the restricted contact length h is small enough to
make the chip utilise the entire tool groove.
Orthogonal Cutting
Tool Groove Size
Tool Groove Style
Work Material
V = 100m/min, ti= 0.2mm, w = 3.0mm
Medium (GT2)
Style B (with standard back wall)
Medium Carbon Steel
Figure 4.5 The effect of tool restricted contact length on chip up-curl and chip breaking (based on [67])
E E
m
T5
3 i
Q.
O
14.0"
12.0"
10.0" *
8.0"
6.0-
4,0-
2,0-
o.o-
Orthogonal Cutting, Groove Style B V=100m/min, ti=0.2mm, w=3.0mm
Medium Carbon Steel
° GT1-large size x GT2-med. size • GT3-small size
T 1 1 0.0 0.1 - 0.2 0.3
Tool Restricted Contact Length, h (mm)
0.4
Figure 4.6 The effect of tool restricted contact length on chip up-curl radius for three chip-breaker groove sizes
(c) Effect of groove styles
A representative photograph of the chip breaking patterns with varying chip-breaker
groove styles is given in Figure 4.7. The chip breaker with a raised back wall (i.e.
Style A in Figure 4.3) has the best chip breakability with only one full turn chip,
while, the chip breaker without backwall (i.e. Style D in Figure 4.3) is the poorest in
chip breakability as shown, requiring more turns in the chips produced, before
breakage occurs.
4.4.1-3 Results of machining with grooved chip breakers based
on 3-D chip flow
Analytical results are derived based on the oblique machining experiments with
different work materials, cutting speeds, depths of cut, feeds, tool nose radii and tool
inclination angles.
Orthogonal Cutting
Tool Groove Size
Work Material
V = 100m/min, ti= 0.2mm, w = 3.0mm
Medium (GT2), h = 0.3mm
Medium Carbon Steel
Figure 4.7 The effect of chip-breaker groove style on chip breaking (based on [67])
77
(a) The effects of depth of cut, d
As seen from Figure 4.8(a), an incremental in depth of cut increases the chip
thickness, and thus improves the chip breakability at low depths of cut, but when
depth of cut is larger than 2 m m , this effect is insignificant Depth of cut has a greater
influence on chip sideflow angle tls. W h e n the depth of cut is less than or around the
tool nose radius, the chip sideflow angle T|s has a very large value, while at larger
depths of cut, there is a very rapid decrease in rjs (see Figure 4.8(b)). Variations of
depth of cut basically have no influence on chip backflow angle % (Figure 4.8(c)).
Tib <de9) ST
10"
o-
-1C
+
X
O
+
X
o
T
+
X
u
— 1 —
+ X
o
1
+
X
o
— 1 — '
o!o 1.0 2.0 3.0 4.0 0.0 1.0 2.0 3.0 4.0 0.0 1.0 2.0 3.0 4.0
Depth of Cut (mm)
(a)
Depth of Cut (mm)
(b)
Depth of Cut (mm)
(c)
Work Material : Medium Carbon Steel
Tool Nose Radius (r) = 1.2 m m
Cutting Speed (V) = 100 m/min
o Feed (f) = 0.04 mm/rev
x Feed (f) = 0.20 mm/rev
+ Feed (f) = 0.40 mm/rev
Figure 4.8 The effects of depth of cut
78
(b) The effects of feed, f
Feed plays a significant role in the chip breaking performance. Increasing the feed
increases chip sideflow angle T)s, chip backflow angle % and chip thickness t2 for all
cutting conditions tested (see a typical set of results in Figure 4.9).
T|b (deg) 30-
Feed (mm/rev)
i r 0.2 0.4 0.6 0.0 0.2 0.4 0.6
Feed (mm/rev) Feed (mm/rev)
Work Material : Medium Carbon Steel o Depth of Cut (d) = 0.5 m m
Tool Nose Radius (r) = 0.8 m m x Depth of Cut (d) = 2.0 m m
Cutting Speed (V) = 100 m/min + Depth of Cut (d) = 4.0 m m
Figure 4.9 The effects of feed
(c) The effects of tool nose radius, r
The effects of tool nose radius on chip breakability is closely associated with the depth
of cut which determines the effectiveness of tool nose radius, i.e., the effect of tool
nose radius is reduced as the depth of cut increases.
A s shown in Figure 4.10(a), the cutting tool with a sharper nose has better chip
breakability due to the larger chip thickness produced. For a given depth of cut, the
79
influence of tool nose radius on chip sideflow angle rjs increases as the feed increases
(Figure 4.10(b)). The larger the tool nose radius, the greater the chip sideflow angle
x\s. The effect of tool nose radius on chip backflow angle TJb is insignificant.
Tib <de9) 30-
l I
0.0 0.2 0.4 0.6 0.0 0.2 0.4 Feed (mm/rev) Feed (mm/rev)
(a) (b)
l r 0.6 0.0 0.2 0.6
Feed (mm/rev)
(c)
Work Material : Medium Carbon Steel o Tool Nose Radius (r) = 0.4 m m
Depth of Cut (d) = 1.0 m m x Tool Nose Radius (r) = 0.8 m m
Cutting Speed (V) = 100 m/min + Tool Nose Radius (r) = 1.2 m m
Figure 4.10 The effects of tool nose radius
(d) The effects of tool inclination angle, i
The relationship between inclination angle i and chip sideflow angle r|s appears to be
approximately linear, as shown in Figure 4.11.
(e) The effects of cutting speed, V
Chip backflow angle % and chip thickness t2 have been shown that they decrease
with the increase in cutting speed (see Figure 4.4). However, chip sideflow angle T|s
appears to be not sensitive to variations of cutting speed V (see Figure 4.12).
30-
20-
10-
Chip Sideflow Angle, T)s (deg) 70
— i —
-10 T
U = 168m/min d = 0.5mm f = 0.2mm/rev r = 0.8mm
Med. Carbon Steel — • 1 1 ~r
5 15 -10 -5 0 Tool Inclination Angle, I (deg)
60
50
40-
30
20-
10-
o i=-6°
• i=+2°
d = 0.5mm f = 0.15mm/rev r = 0.8mm Med. Carbon Steel
1 — i — ! — i — i — | — i — | — i — | — i — r
0 50 100 150 200 250 300 350 Cutting Speed, V (m/min)
Figure 4.11 The effect of inclination angle on chip sideflow
Figure 4.12 The effect of cutting speed on chip sideflow
(f) The effects of work material
In general, higher carbon steel produces thinner chips which are more difficult to
break (see Figure 4.13). Chip thickness variations are significant at high feeds.
However, the chip sideflow angle T|s and chip backflow angle T\b are not sensitive to
variations of the work materials tested (low, medium & high carbon steels).
4.4.2 Analysis of Grooved Chip Breakers and
Power Consumption Rate
For a given set of cutting conditions, the minimum power consumption for grooved
chip breakers can be achieved by selecting the appropriate restricted contact length h,
groove size and groove style.
81
0.0 0.2 0.4 0.6 0.0 0.2
Feed (mm/rev) Feed (mm/rev)
i i
0.2 0.4 0.6 Feed (mm/rev)
Tool Nose Radius=1.2 m m Depth of Cut=3.0 m m Cutting Speed=100 m/min
o Low Carbon Steel (C 0.25% Mn 0.80% Si 0.40% HB = 140)
x Medium Carbon Steel (C 0.40% Mn 0.75% Si 0.27% HB = 150)
+ High Carbon Steel (C 1.0% Cr 1.25% HB = 225)
Figure 4.13 The effects of work material (carbon steels)
Machining experiments with a secondary rake restricted contact tool were conducted to
study the effect of tool/chip contact length on power consumption under a fairly wide
range of undeformed chip thickness ti, tool/chip contact length he, rake angle a and
cutting speed V. Figure 4.14 gives a representative result. At a given ti value, an
increase in the tool/chip contact length causes the increase in cutting force within the
restricted contact region, while the cutting forces remain constant with the increase in
contact length within the natural contact region. Experiments at varying cutting speeds
and tool rake angles have shown this effect consistently, indicating that the reduced
contact length within the restricted contact region is the key to reduced power
consumption.
82
• \y= 0.071 mm n ti = 0.113mm A ti= 0.141mm o t1= 0.177mm • 11=0.198mm
Rake Angle : -5 ° Cutting Speed:
100 m/min Width of Cut: 3 m m Medium Carbon Steel
Figure 4.14 The effect of tool/chip contact length on power consumption
However, when machining with a grooved chip-breaker tool, the power consumption
rate is different. As shown in Figure 4.15 for a set of typical specific cutting pressure
Ps - restricted contact length h relationships, there exist minimum power regions for
all three tool groove sizes, depending on the tool groove utilisation rate. The restricted
contact length h is a major factor in determining tool groove utilisation due to its
significant effect on chip backflow angle %. For all three tool groove sizes, the
differences in power consumption among all the groove styles (A, B, C and D) are
very explicit and this effect can be explained in terms of the geometry of the grooves
and the corresponding utilisation of these grooves by the chip.
^ 2 . 0 H z w 1.8-
£ . 1.6-
U.
» 1.2-
o 1.0-
S 0.8 H to
0.6
restricted contact region
p e e° e » — * ~
natural contact region • • i • •
0.00 0.30 0.60 0.90 1.20 Tool/Chip Contact Length, h c (mm)
83
Specific Cutting Pressure Ps (10 MPa)
2.3-
2.2-
2.1-
2.C
1.9
2.2-
"T 1.8
GT2, t!=0.2mm
0.15 0.25 0.35 0.1 h (mm)
-1 1
0.2 0.3 0.4 0.05 h (mm)
0.15 0.25 h (mm)
Work Material : Medium Carbon Steel
Width of Cut (w) = 3.0 m m
Rake Angle (a) = 0°
Cutting Speed (V) = 100 m/min
0 Groove Style A
• Groove Style B
+ Groove Style C
x Groove Style D
Figure 4. 15 The effect of tool restricted contact length on specific cutting pressure in machining with grooved chip breakers
4.5 ANALYSIS OF 3-D CHIP FLOW AND CHIP
CURLING WITH GROOVED CHIP BREAKERS
4.5.1 Effect of the Chip Backflow Angle, r)b
Chip backflow angle % plays a key role in tool groove utilisation, which determines
the chip breakability and power consumption rates. Figure 4.16 shows a typical chip
curling pattern in the case of a grooved chip breaker with restricted contact length (i.e.
h < h n ) . It is apparent from the figure that the smaller the chip curl radius ru , the
greater is the tool groove utilisation. The chip may fully utilise the groove profile
when the chip curl radius ru is equal to the groove radius Ro,
ru = rumin = Ro = B/(2sin0) (4.2)
where 8 is the groove tangent angle which has a very close relationship with the chip
backflow angle %. From the relationships shown in Figure 4.16, it can be derived
that if (Tlb-a) < 6, the chip can not fully utilise the tool groove profile; while if (%-
a) > 0, the chip will "overuse" the groove profile, which would unnecessarily
consume more power due to the increased friction between the chip and the tool
groove surface. However, in designing, the ratio (T|-a)/9 = 1.2 may be taken as a
reasonable basis to guarantee the best utilisation of the tool groove profile.
It has been shown that the chip up-curl radius (ru) varies within a chip breaking cycle
from the use of high speed filming techniques for the actual chip breaking process
[79]. This variation results in the variation of tool groove utilisation. Therefore the
term "full utilisation of the groove" has to be interpreted as the "maximum possible
utilisation of the chip breaker groove".
Figure 4.16 The chip backflow effect on tool groove utilisation
4.5.2 Effect of the Chip Sideflow Angle r)s
In three-dimensional oblique cutting, the effectiveness of grooved chip breakers is
reduced due to the associated chip sideflow angle t|s. A s shown in Figure 4.17, when
chip sideflow angle rjs is not equal to zero, the equivalent tool restricted contact length
he and the equivalent groove width B e become larger, i.e.,
he = h/cosr\s
B e = B/cosTis (4.3)
Figure 4.17 The chip sideflow effect on the effective tool groove parameters
The groove profile will be of no use if the equivalent restricted contact length hg is
larger than the tool natural contact length hn, and when r|s is larger than, say, 75
degrees, he and B e will go up rapidly with a small increase in T|s. In this case, the
direction of chip sideflow will be nearly parallel to the cutting edge, therefore a
standard grooved chip breaker is no longer useable for chip curling.
4.6. THE ESTABLISHMENT OF A KNOWLEDGE BASE
The knowledge base is used for predicting the tool/chip natural contact length hn, chip
sideflow angle r|s, chip backflow angle r|b and for selecting the groove parameters for
effective chip breaking and minimum power consumption.
4.6.1 Knowledge Rules for Predicting the Natural Contact Length h„
The groove effect of a restricted contact tool is operational only when the tool contact
length h is smaller than the natural tool/chip contact length hn, for a given set of
cutting conditions. The h n value can be determined by the experimental T|b-h
relationship diagram for orthogonal cutting conditions or rtb-he relationship diagram
for oblique cutting conditions. Figure 4.18 shows a typical diagram that can be used
for deterrnining h n.
Figure 4.18 Determination of tool/chip natural contact length hn
O n the basis of 3-D Chip R o w Database, hn values can be estimated directly for the
combinations of 3 work materials, 3 tool nose radii, 5 depths of cut and 6 feeds with
cutting speed V = lOOm/min and tool rake angle a = -5°. The effects of V and a can
be derived by knowledge rules based upon the Reference Database.
(a) Rules for the effect of V
The h„ value increases with the increase in cutting speed V for all the cutting
conditions tested. Given below is a representative rule for the effect of V on hn with
V = 100 m/min as the standard value for comparison.
IF 150 <V < 180 m/min
T H E N the natural tool/chip contact length hn will have a
1 5 % increase (AhnV = +15%) at low feeds (f < 0.15mm/rev) or
1 8 % increase (AhnV = +18%) at medium feeds (0.15 < f < 0.3mm/rev) or
2 2 % increase (AhnV = +22%) at high feeds (f l> 0.3mm/rev).
(b) Rules for the effect of a
The effect of a on hn is very little. In general, the hn value will have a slight
decrease with the increase in a. A n example of the rules for the effect of a on hn is
given below with a = -5° as the standard value for comparison.
IF 0<a <5°
T H E N the natural tool/chip contact length will have a
9% decrease (Ahna = -9%) at low feeds (f < 0.15mmA-ev) or
7 % decrease (Ahna = -7%) at medium feeds (0.15 < f < 0.3mm/rev) or
5 % decrease (Ahna = -5%) at high feeds (f > 0.3mm/rev).
4.6.2 Knowledge Rules for Predicting the Chip Backflow Angle r|b
The restricted contact length h, tool rake angle a, feed f (or undeformed chip
thickness ti) and cutting speed V have significant influence on chip backflow angle
% •
(a) Rules for the effect of h :
The chip backflow angle t|b is sensitive to small variations of h values. Knowledge
rules have been derived to predict r)b for all possible h values under various cutting
conditions based on the Reference Database. The standard h value is different at
different machining conditions. One of these rules is shown below.
IF 0.15 < h < 0.19 mm, a = 0°, V = 100 m/min
T H E N Chip backflow angle Tjb will have a
4 0 % increase (Ar]bh = +40%) at low feeds (f < 0.15 mm/rev) or
3 8 % increase (Artbh = +38%) at medium feeds (0.15 < f < 0.3mm/rev) or
2 8 % increase (Arjbh = +28%) at high feeds (f > 0.3 mm/rev)
comparing with that at the standard value h = 0.226 m m .
(b) Rules for the effect of a :
Knowledge rules for the effect of rake angle a on chip backflow angle T]b are
summarised based on the comparison with the standard rake angle value, a = 0
degree, at different machining conditions. Given below is an example.
IF 8° < a < 10°, f <, 0.15 mm/rev, h < hn
THEN Chip backflow angle n,b will have a
3 % increase (Arj^ =+3%) at low cutting speeds (V ^ lOOm/min) or
8 % increase (Arj^ = + 8 % ) at med. cutting speeds (100 < V < 200m/min) or
1 2 % increase ( A n ^ =+12%) at high cutting speeds (200 < V < 300m/min) or
1 8 % increase (Arj^ =+18%) at very high cutting speeds (V > 300m/min).
(c) Rules for the effect of cutting speed V
Taking cutting speed V = lOOm/min as the standard value for comparison, knowledge
rules for the effect of V on T]b are developed. A representative rule is shown as
follows:
IF 200 <V <250m/min, f <0.15mm/rev, h <hn
T H E N Chip backflow angle Tjb will have a
1 0 % decrease (AT|bV
1 4 % decrease (Ar|bv
1 6 % decrease (ArjbV
2 0 % decrease (AT|bV
3 2 % decrease (Anb V
5 0 % decrease (Ar|bV
5 8 % decrease (AT]bV
6 5 % decrease (ArjbV
= -10%)
= -14%)
= -16%)
= -20%)
= -32%)
= -50%)
= -58%)
= -65%)
at
at
at
at
at
at
at
at
a >10°
5° < a < 10°
2°<a <5°
-2°<a <2°
-5°<a <-2°
-8°<a <-5°
-15° < a <-8°
a <-15°-
or
or
or
or
or
or
or
(d) Rules for the effect of feed f
The feed rate has a significant effect on chip backflow angle T|b. Six feed values
(0.04, 0.1, 0.2, 0.3, 0.4, 0.5 mm/rev) are selected as the standard values in the
established 3-D Chip Flow Database. A typical rule for the effect of feed on T]b is as
follows:
IF W o r k material
Tool nose radius
Depth of cut
While feed value
is
is
is
is
low carbon steel
r = 0.8mm
1.0 < d < 2.0
0.2<f<0.25mm/rev
and
and
and
THEN Chip backflow angle will have a 30% increase (Ar|bf = +30%)
comparing with that at the standard feed value f = 0.2mm/rev
in the 3-D Chip Flow Database.
4.6.3 Knowledge Rules for Predicting the Chip Sideflow Angle rjs
The tool nose radius r, tool inclination angle i, depth of cut d and feed f are the
major factors influencing the chip sideflow angle Tls in oblique machining.
(a) Rules for the effect of tool nose radius r
The tool nose radius r plays a significant role in determining chip sideflow angle rjs.
Given below is a sample of such rules for its effect on Tis.
IF Work material
Feed
Depth of cut
While tool nose radi
is
is
is
ius is
high carbon steel
0.3 < f < 0.4mm/rev
d < 0.5mm
1.4 <r< 1.6mm
and
and
and
THEN Chip sideflow angle rjs will have a 25% increase (A^ =+25%)
comparing with that at the standard tool nose radius value r = 1.2mm
in the 3-D Chip Flow Database.
(b) Rules for the effect of tool inclination angle i
In formulating the knowledge rules for the effect of inclination angle, i = 0° is taken as
the standard value for comparison, an example of such rules is shown as follows :
IF
T H E N
Work material is medium carbon steel and i = ii > 0°
Chip sideflow angle T|s will have an increment T|sj = 1.15ii
(c) Rules for the effect of depth of cut d
The depth of cut has the most significant influence on chip sideflow angle TJS. Five
depth of cut values (0.5, 1.0, 2.0, 3.0, 4.0 m m ) are selected as the standard values.
One of the knowledge rules for depth of cut is given below:
IF Work material
Feed
Tool nose radius
While depth of cut
is
is
is
is
medium carbon steel
0.1 <f <0.2mm/rev
r = 1.2mm
1.2 <d < 1.5mm
and
and
and
T H E N Chip sideflow angle T|s will have a 2 2 % decrease (Arj^ = -22%)
comparing with that at the standard value d = 1.0mm in the 3-D
Chip Flow Database.
(d) Rules for the effect of feed f
Knowledge rules for the effect of feed f on chip sideflow angle rjs are derived based
on the selected six standard feed values. Given below is an example of the rules used:
IF Work material
Tool nose radius
Depth of cut
While feed
is
is
is
is
low carbon steel
r = 0.4mm
0.8 < d < 1.2mm
0.35<f<0.4mm/rev
and
and
and
T H E N Chip sideflow angle rjs will have an 18% decrease (Arjsf = -18%)
comparing with that at the standard value f = 0.4mrn/rev in the 3-D Chip
Flow Database.
4.6.4 Knowledge Rules for Determining the Minimum
Power Consumption
(a) Rules for setting up the restricted contact length - power consumption relationship
In the experiments for setting up the Grooved Chip Breaker Database, six standard h
values (0.36, 0.30, 0.24, 0.18,0.12 & 0.06 m m ) were used at V = 100 m/min, a =
0°. The effect of different h values on power consumption under various cutting
conditions can be summarised into knowledge rules on the basis of the Reference
Database. One example of such rules is given below.
IF 0.3 < h < 0.33 m m at high feed value (f 0.3 mm/rev)
T H E N power consumption will have a 3 % increase (APh = + 3 % ) comparing with
that at standard value h = 0.3mm in the Grooved Chip Breaker Database.
(b) Rules for determining the h values for minimum power consumption
The restricted contact length h is the most important parameter for grooved chip
breakers. Generally speaking, the smaller the h value, the lower the power
consumption. However, a too small h value may give rise to an adverse effect. The
reason for this is that a decrease in h may cause an increase in T]b and if the Ti.b value
is excessive it will consume more power. Based upon the established Grooved Chip
Breaker Database, the rules for determining h values for minimum power
consumption are summarised. Given below is one of the samples.
IF tool groove size is large (GT1) and groove style is A
T H E N power consumption is minimum for
h = 0.30 m m at medium feeds (0.15 < f < 0.3mm/rev) or
h = 0.36 m m at high feeds (f > 0.3mm/rev).
4.6.5 Knowledge Rules for Selecting the Effective Groove Profile
(a) Rules for the selection of tool groove sizes (GT1, GT2 & GT3)
(i) IF feed value is low (i.e. f < 0.15mm/rev)
T H E N small groove size (GT3) must be selected with h = 0.5hn.
(ii) IF feed value is medium (i.e. 0.15 < f < 0.3mm/rev)
T H E N medium groove size (GT2) is selected with h =0.6hn or
small groove size (GT3) is selected with h =0.7hn.
(hi) IF feed value is high (i.e. f > 0.3 m m )
T H E N large groove size (GT1) is selected with h =0.7hn or
medium groove size (GT2) is selected with h = 0.8hn.
(b) Rules for the selection of tool groove styles (A, B,C &D)
(i) IF all the cutting conditions are same
T H E N the chip breakability is in decreasing order for tool groove styles
(from A to D ) for any given tool groove size.
(ii) IF feed value is low (f £ 0.15mm/rev), the chip breaking is usually a
problem.
T H E N the preferential selection of tool groove styles is from A to D.
(iii) IF feed value is high (f ^ 0.3mm/rev), the chip breaking is usually not a
problem.
T H E N the tool groove style with minimum power consumption is preferred.
(iv) IF feed value is medium (0.15 < f < 0.3mm/rev), the integrated effects
of chip breakability and power consumption should be considered.
T H E N when the difference in power consumption is less than 25%,
the tool groove style with better chip breakability is preferred.
4.7 A STRATEGY FOR DESIGNING EFFECTIVE GROOVED CHIP BREAKERS
The design of an effective grooved chip breaker greatly depends on the optimum
groove utilisation, i.e. effective chip breaking and minimum power consumption.
4.7.1 Determining the Optimum Restricted Contact Length h
As discussed before, an appropriate selection of h value is very important. A very
small h value will weaken the strength of the tool cutting edge, furthermore, it might
consume more power when machining with the specific grooved chip breakers,
mainly due to the excessive chip backflow angle T|b. Figure 4.19(a) shows a typical
T|b - h relationship from the experiments, when h < hn, the chip will flow towards
the groove with an effective chip backflow angle T|be and h^ is the assumed h value
from the point of view of effective chip breaking. Figure 4.19(b) shows another
typical Ps - h relationship from the experiments, where there exists a specific h
value (hp) at which the power consumption reaches its minimum value. In the region
where h > hp, the power consumption reduces with the decrease in h, while in the
region for h < hp, the effect is opposite. The curve in this region is not as steep as
that in the region for h > hp. Therefore, the optimum h value is determined by the
following relationship (hn and hp are determined by the associated rules),
h = {min( h^, hp )}cosrjs (4.4)
where the chip sideflow angle rjs is predicted based on the 3-D Chip Flow Database
and the .associated knowledge rules. If w e assume TJS sttj is the standard chip sideflow
angle in the 3-D Chip Flow Database, Tls can be determined by the following
equation,
Hs = Tlsi + (1 + ATisr)( 1 + A T ] ^ )(1 + Arisf)risstd (4.5)
Tib 4
TJbe
Psi
se .Z*
(a) Typical T)D-h relationship (b) Typical Ps-h relationship
Figure 4.19 Deterrnining the optimum restricted contact length
4.7.2 Determining the Effective Groove Parameters
From the selected restricted contact lengths h associated with the input cutting
conditions, the chip backflow angle T]b can be predicted using the knowledge rules
and the established 3-D Chip H o w Database. If n.b std is assumed to be the standard
chip backflow angle in the 3-D Chip Flow Database, the predicted chip backflow
angle is determined as follows :
% = (1 + ATibh)( 1 + AribvX 1 + AT]ba)( 1 + ATibf)Tib std (4.6)
On the basis of this predicted chip backflow angle rjb, the groove parameters (refer to
Figure 4.3) are calculated by the following set of equations :
Groove Tangent Angle (6):
8 = (Tib-a)/1.2 (4.7)
Groove Width (B):
2.0 mm, for large groove size (GT1)
B = 11.5 m m , for medium groove size (GT2) (4.8)
1.0 m m , for small groove size (GT3)
Groove Radius (Ro) :
Ro = B/2sin8 (4.9)
Groove Depth (d) :
d = Ro-[R02-(B/2)2] 1/2 (4.10)
Raised Height of Groove Back Wall (di) :
di = 0.5d (4.11)
Reduced Height of Groove Back Wall ((fc) :
d2 = 0.3d (4.12)
4.7.3 Schematic Diagram for Designing Effective
Grooved Chip Breakers
On the basis of the established knowledge-based system for designing effective
grooved chip breakers, a schematic diagram for determining the most effective tool
restricted contact length h as well as the groove parameters has been prepared and is
shown in Figure 4.20. This system has been developed by using Turbo P R O L O G
language.
INPUT Cutting Conditions (V, d & f), Tool Geometries (r, a & i) and Work Material
1 PREDICT Chip Sideflow Angle : T|s = T|si + (1 +Ar|sr) (1 +AT]sd) (1 +ATlsf)T|s std
PREDICT Natural Tool/Chip Contact Length hn
I SELECT Tool Groove Style (GT1, GT2 & GT3) and h^ based on Effective Chip Breaking
DECIDE Tool Groove Style (A, B, C & D) and hp .based on Minimum Power Consumption
I ~ DETERMINE Optimum Restricted Contact Length : h = { min( h^, hp) } CCS(r|s)
I PREDICT Chip Backflow Angle : l\h = (l+ATlbh)(l+ATlbV)(l+ATlba)(l+ATlbf)rtb std
. L . DETERMINE Chip Breaker Groove Tangent Angle: 8 = ( rjb - a )/ 1.2
j ; CALCULATE GROOVE PARAMETERS
(1) Groove Width B (2) Groove Radius R0
(3) Groove Depth d (4) Raised Height of Groove Back Wall d2
(5) Reduced Height of Groove Back Wall dj
i OUTP U T The Most Effective h Value and Groove Parameters
Figure 4.20 Flow chart for the knowledge-based system developed
4.8 SUMMARY AND CONCLUDING REMARKS
(1) Most commercially available cutting tools are designed on the basis of the
traditional "try and see" methods which very often do not provide the best
results. Therefore, it is imperative to explore a more scientific design strategy in
order to improve the performance of cutting tools in machining.
(2) An alternative to the traditional design of chip breakers is presented in this
chapter in the form of knowledge-based system incorporating a well established
database system including Reference Database, Grooved Chip Breaker Database
and 3-D Chip Flow Database.
(3) In the present work, conventional grooved chip breakers are selected as the
typical tools since the analysis of them could be regarded as a suitable basis for
the general chip breaker design. The combined effects of tool restricted contact
and groove configurations have been studied systematically in terms of three-
dimensional chip flow. It has been shown that efficient chip breaking at the
minimum power consumption can be achieved through the optimum groove
utilisation of the chip breakers.
(4) Although the knowledge-based system is only for designing effective grooved
chip breakers, the methodology described in this chapter could be extended to the
design of a more complicated toolface configuration.
100
CHAPTER 5
COMPREHENSIVE TOOL WEAR ESTIMATION IN FINISH-MACHINING BY MULTIVARIATE
TIME SERIES ANALYSIS
5.1 INTRODUCTION
In fimsh-macriining, the traditional (major) flank or crater wear estimation alone is no
longer adequate and the wear on the minor flank face, such as minor flank wear or
groove wear at the minor cutting edge, is of more importance since it directly affects
the surface quality and dimensional accuracy of a finished product. It is obviously
desirable for these wear states to be effectively monitored as well, because it is likely
that wear in these areas may reach critical points earlier than those in the major flank
and crater, such that the optimum cutting conditions or tool change policy in a finish-
machining has to be set based on these wear types. Therefore, a more comprehensive
monitoring strategy involving multi-sensor or multi-modelling is called for in order to
estimate more than one type of tool wear, or comprehensive tool wear estimation.
Employing multi-sensor or multi-modelling strategies has been identified in a recent
survey conducted for CIRP [5] as one of the three promising directions in machining
process monitoring and control research. Interesting work has been reported in
integrating force and acoustic emission (AE) signals via neural networks [11-12, 80].
Chryssolouris [37] evaluated the effectiveness of sensor integration for tool wear
estimation by neural network, least-squares regression, and the group method of data
handling ( G M D H ) algorithm using simulation data. Both papers reported better
estimation of flank wear by integrating multi-sensory information than by using a
single sensor. For finish-machining where more than one quantity is to be estimated,
however, a multi-sensor and multi-modelling strategy as suggested in [5] becomes
necessary.
101
The "comprehensive" monitoring strategy has been addressed less frequently,
perhaps because of the complexity of the machining process. If more than one
quantity is to be estimated, more complexity will be encountered. This places higher
demands on signal processing and analysis techniques which shall be able to "single
out", from the signals, particular ingredients sensitive to particular quantities to be
estimated. Otherwise, multi-sensory techniques will do more harm than help. The
spectrum analysis is a technique commonly used to single out frequency components
to be correlated to tool wear [81-83]. Time domain methods, such as using
autocorrelation coefficients of cutting force signals have been reported [13]. It,
however, has been recognised that the cutting process is a stochastic process due to
the existence of inevitable material property variations and other uncertainties. The
necessity of employing stochastic analysis for cutting dynamics was emphasised in
[84]. Interesting work on correlating coefficients of Autoregressive (AR) models of
A E signals to flank wear was reported [7], though appreciation of the results is
impaired by inadequate physical interpretations. Another example of using stochastic
analysis is to detect tool breakage by monitoring the residuals of an A R model
obtained from cutting torque signals [27]. The residual analysis has been proven to be
very effective to detect abrupt changes in the cutting process, such as tool breakage.
Since groove formation at the minor cutting edge occurs only under certain cutting
conditions, in this chapter, minor flank wear is first estimated along with major flank
and crater wear, aiming at developing a comprehensive estimation strategy for tool
wear in finish-machining. As using 3-D dynamic cutting force is an effective means
for providing necessary information about wear states at different tool faces, the
estimation strategy is based on the cutting force measured in terms of its three
orthogonal components, from which trivariate Autoregressive Moving Average Vector
( A R M A V ) models [35] were developed. The dispersion analysis (DA) [35, 85] based
on A R M A V models led to the discrimination between various modes of force
102
variations in a quantitative way, such that correlating them to various quantities to be
estimated was made possible. The correlation results were supported by physical
interpretations.
5.2 TOOL WEAR EXPERIMENTS
5.2.1 Description of Experiments
Dynamic components of the cutting force were measured by using a tool dynamometer
(KISTLER 9257A). Figure 5.1 shows the experiment setup used. The data
acquisition system used in the experiment is a Macintosh-based 12-bit A/D converter
(MacADIOS JJ with Maclnstrument software).
Workpiece
Low-Pass " Filters ]
— »
Charge Amplifiers
Tool
Multichannel Data Acquisition System
r-
O KISTLER Dynamometer
»
• Computer
Figure 5.1 Experiment setup
Four typical cutting conditions are selected for tool wear experiments as shown in
Table 5.1. As w e are only interested in the development patterns of particular types of
tool wear, the degraded tool tests, i.e. using relatively softer tools as recommended in
[86], are adopted to shorten the time-consuming and costly tool wear experiments.
103
Table 5.1 Machining conditions used in tool wear experiments
Machine Tool
Tool Insert Type
Tool Material
Tool Geometry
Work Material
Workpiece Size
Cutting
Conditions
Cutting Fluid
Colchester Mascot 1600 (9.3 K W )
TNMA160408F (flat-faced tool)
Carbide : Grade 883, S E C O
0°, 5°, -6°, 90°, 60°, 0.8
AISI4140 (BHN=275-320)
Length =lm and Diameter =100mm
1. V=l 15 m/min f=0.1 mm/rev d=0.5mm
2. V=145 m/min f=0.1 mm/rev d=0.5mm
3. V=145 m/min f=0.06 mm/rev d=0.5mm
4. V=145 m/min f=0.06 mm/rev d=0.25
No
A typical record of the dynamic cutting force is illustrated in Figure 5.2 in terms of its
three orthogonal components, i.e., Fx, Fy and Fz.
In order to assure that the experimental conditions are as close as possible to practical
machining operations, the machining process was interrupted periodically with an
increment in period of about 5 minutes under cutting condition Group 1, and 2.5
minutes under Groups 2-4. The tool was replaced by a fresh one at each interruption
such that every tool remained in thermal continuity until it was replaced. Just before
each tool replacement, a set of 524 data points was sampled for each channel.
Therefore, the experimental results consist of 8 tools and 8 sets of data from each
channel under Group 1, and 7 tools and 7 sets of data from each channel under
Groups 2-4. Before the dynamic cutting force in terms of its three orthogonal
components was sampled into a multi-channel data acquisition system with a sample
interval equal to 60 ps (about 16.7 KHz), low-pass filters with a cut-off frequency of
4 K H z were applied, considering the 4 K H z resonant frequency of the dynamometer.
-50 » - '
10 15 20 25
Time (ms) (a) Dynamic Cutting Force in Feed Direction
30
Time (ms) (b) Dynamic Cutting Force in Thrust Direction
Time (ms)
(c) Dynamic Cutting Force in Main Cutting Direction
Figure 5.2 Dynamic cutting force measured in terms of its three orthogonal components
105
5.2.2 Definition of Comprehensive Tool Wear Parameters
Eight parameters were selected to describe the tool wear states as shown in Figure
5.3, primarily in accordance with CIRP tool wear terminology [87]. The eight tool
wear parameters are roughly classified into three categories with respect to different
tool faces, i.e., the major flank area (VB, K S & V G ) , crater area (KT, K B & K K )
and minor flank area (VB' & N ) .
Plane P m
S : the point on the original major cutting edge at the middle of depth of cut
P : the plane perpendicular to the major cutting edge through Point S
VB
KS
KB
major flank wear
retract of the cutting edge
crater width on the rake face
VB' : minor flank wear
VG
KT
K K
N :
length of the groove (notch)
crater depth on the rake face
crater length on the race face
nose wear
Figure 5.3 Definition of comprehensive tool wear parameters (based on [87])
106
5.2.3 Tool Wear Measurement and Wear Development Patterns
The scanning electron microscope (SEM), stereo microscope with camera attachment,
surfcom (a surface roughness measuring instrument) and coordinate measuring
machine (CMM) were used jointly to measure wear parameters selected.
Table 5.2 gives the measurement results of VB, KT and VB1 for cutting condition
Group 4. Table 5.3 tabulates the measurement results of the all eight types of wear
for cutting condition Groups 1-3 and the developments of these types of tool wear
were plotted in Figures 5.4 and 5.5. A typical group of tool wear photographs taken
by S E M is shown in Figure 5.6, indicating the wear situations on different tool faces.
The formation of groove wear at the minor cutting edge (see Figure 5.6(d)) was
noticed but not selected for investigation because of its inconsistent appearance under
the machining conditions used in the experiment and it will be addressed separately in
the following chapter.
Table 5.2 Tool wear measurement result for cutting condition Group 4
Time (min)
VB (mm)
KT (pm)
VB' (pm)
0
0
0
0
2.45
0.15
23.9
15.1
5.1
0.19
37.2
17.5
7.6
0.26
48.1
58.3
10
0.46
53.8
99.5
12.4
0.50
108
186
15
0
119
19
1
Table 5.3 Tool wear measurement results for cutting condition Groups 1-3
Tool Wear Measurement for Cutting Condition Group 1
Time
(min)
0
2.67
5
10
15
20
25
34
VB (mm)
0
0.12
0.16
0.26
0.34
0.58
0.64
0.73
KS (pm)
0
12.8
25.7
47.1
80.0
104
127
143
VG (mm)
0
0.09
0.10
0.40
0.70
0.75
0.80
0.84
KT (pm)
0
22.2
33.3
38.8
66.7
167
206
222
KB (mm)
0
1.1 1.2
1.2 1.2 1.4 1.4
1.5
KK (mm)
0
0.4
0.60
0.63
0.64
0.66
0.68
0.70
VB' (pm)
0
15.3
28.6
71.4
186
229
236
243
N
(pm)
0
7.1
21.4
32.8
54.3
91.4
113
136
Tool Wear Measurement for Cutting Condition Group 2
Time
(min)
0
2.58
5.08
7.5
10
12.42
15.16
VB
(mm)
0
0.17
0.23
0.30
0.48
0.57
0.80
KS (pm)
0
0.30
83.3
100
108
150
200
VG
(mm)
0
0.20
0.26
0.45
0.69
0.73
0.94
KT (pm)
0
25.2
38.3
48.3
63.2
104
144
KB (mm)
0 1.0 1.05
1.15
1.20
1.30
1.40
KK (mm)
0
0.48
0.64
0.66
0.70
0.72
0.76
VB' (pm)
0
10.0
18.6
69.3
101
203
211
N
(pm)
0
5.2
16.5
34.1
72.4
101
105
Tool Wear Measurement for Cutting Condition Group 3
Time (min)
0
2.5
5
7.67
10.08
12.5
15
VB
(mm)
0
0.15
0.20
0.28
0.46
0.52
0.74
KS (pm)
0
28.3
81.7
100
103
153
166
VG
(mm)
0
0.17
0.21
0.44
0.64
0.74
0.88
KT (pm)
0
24.8
35.7
47.2
55.6
114
131
KB (mm)
0 0.9
1.0
1.1
1.1
1.2
1.3
KK (mm)
0
0.46
0.58
0.60
0.62
0.65
0.69
VB' (pm)
0
33.3
36.7
83.3
117
197
202
N
(pm)
0
5.6
15.6
35.3
68.2
93.5
96.7
108
VB(mm) (a) Major Flank Area KS(um)
10 15 20 25 30
KB* KK(mmz) (b) Crater Area KT(pm) 250
VB'(pm) (c) Minor Flank Area N(pm)
10 15 20 25 30 Time (minutes)
Figure 5.4 Tool wear development for cutting condition Group 1 (V=l 15 m/min, f=0.1 mm/rev, d=0.5 m m )
109
VB(mm)
0.80-
0.60-
0.40"
0.20-
0.00
KS(um)
"200
-150
VB(mm) 0.8
0.6-
(a)
100 0.4-
- 50 0.2 -
KS(um)
200
KB* KK(mnr) 1.2
VB'(um) 250
(C)
KT(um) KB* KK(mmz) 150 1.0
100
N(pm) VB'(pm) 120 250
(C)
5 10 15
Time (minute)
— Group 2 —
KT(um)
150
100
N(pm)
100
5 10 15
Time (minute)
— Group 3 —
Figure 5.5 Tool wear development for cutting condition Groups 2 & 3
1 1 0
I 1 200 pm I 1 200 pm
(a) major flank wear (b) crater wear and nose wear
I 1 100 pm I 1 100 pm
(c) minor flank wear (d) groove wear at the minor cutting edge
Figure 5.6 Tool wear observed under a scanning electron microscope (SEM)
5.3 MULTIVARIATE TIME SERIES TECHNIQUES
5.3.1 Trivariate ARMAV Models
It is known that the dynamic cutting force, which is the variation from the average
cutting force, contains richer information about tool/workpiece interactions during
machining than the latter alone [84]. It has been shown that the dynamic cutting force
111
is a stochastic signal which roughly obeys the normal distribution [88]. It is also
appropriate to regard the dynamic force as stationary processes at different stages of
wear development, because it takes only a fraction of a second for a set of a few
hundred data points to be sampled each time. In summary, it is appropriate to apply
statistical methods for stationary normal processes to the dynamic cutting force signal.
As a way of analysing the dynamics in the cutting force measurements, trivariate time
series models, developed from the data, are used, since they give a concise parametric
representation of the signals.
When a dynamic process represented by its p-components is sampled at uniform
intervals, A, the resulting discrete series of observation vectors, Xt, t=l, 2,..., N;
can be represented by
n m x t = 5 > k X t . k + at-]>>kat.k (5.1)
k=l k=l
where the p-dimensional vector of process variables is given by the observation
vectors X t = [Xlt, X2t,..., XptJT, and the white noise at= [alt, a2t,..., apj
1", with the
properties of E[aJ = 0 and E[2kt&tjj^] = 8 ^ . Superscript T denotes vector transpose,
E expectation, 6k the Kronecker delta function and o a the covariance of at.
The model in Equation 5.1 is termed as Autoregressive Moving Average Vector model
of autoregressive order n and moving average order m denoted by A R M A V ( n , m ) .
The parameter matrices <X>k's and 6k's are estimated based on the observation vectors.
The orders of an adequate model can be determined by the F-test [35] or by the
Akaike's Information Criterion (AIC) [89]. It can be shown that an A R M A V ( n , m )
model can be approximated by an autoregressive model, i.e., ARV(n) of a sufficiently
high order [35]. A n A R V model requires much less computations than an A R M A V
1 1 2
model does, such that it is more attractive for on-line implementations. When applied
to the oblique machining process, an ARV(n) model has a much simpler form:
X t = £ < D k X t . k + a t (5.2) k=i
T where X t = ( X l t ,X 2 t,X 3 t) ,
/
ok= *llk °12k °13k
T °2ik °22k °23k | and at=(ait,a2t, a3t)
V °31k *32k °33k
Such a model expresses the observed trivariate series, X l t = Fxt = feed force, X 2 t =
Fyt = thrust force, and X 3 t = Fzt = main cutting force, as linear combinations of past
observation vectors Xt.k plus the white noise at and therefore describes the
instantaneous dynamics of the cutting process.
5.3.2 Dispersion Analysis
Once an adequate ARMAV model is determined, the dispersion analysis (DA) can be
introduced to make discrimination among various modes of dynamic force variations
in a quantitative way. It can be shown [35] that the correlation matrix of the
measured variables is a weighted linear combination of the eigenvalues, X-v (i=l,2,
..., n, for each series) as follows,
113
Yk=E[XtxJt] = XdiXi (5.3) i=l
If k = 0, one obtains the process variance for the measured variables as
Y o=E[XtxS = |;di (5.4) i=l
where dj is the dispersion associated with eigenvalue X^ and given by
d i = g i Z — r — (5.5) k=i 1 - X,jX,k
and gj is calculated by the following equation,
gi=
n-l
n < h-K) fc=l,faei
(5.6)
The dispersion percentage, Dj, describes the contribution of the roots or ultimately the
frequencies in the series to the process variation 7o and given as :
D i = - (5.7) Yo
1 1 4
In this way the process variation y0 is decomposed into contributions of process
eigenvalues in terms of dispersion (D4) quantitatively. O f particular interest are the
dispersion percentages (Dj) associated with eigenvalues (X*) occurring in complex
conjugate pairs (Xi and X^) which describe the oscillating variation of the machining
process. The frequency corresponding to a pair of complex conjugate eigenvalues is
given by [35],
f n (Hz) = 27tA VI ln^j X2)
-,2 r
COS 1/ ^1+^2 \
i4x^x~2 (5.8)
where A is the sample interval in seconds.
The significance of dispersion analysis lies in that the relative importance of the
oscillating mode of each existing frequency can be established such that analysis and
interpretation in terms of physical phenomena, such as natural frequencies of the
tool/tool holder system and machine tool structural frequencies, can be carried out in a
quantitative manner.
5.4. TOOL WEAR ESTIMATION BY DISPERSION
ANALYSIS BASED ON ARV MODELS
5.4.1 ARV Modelling of 3-D Dynamic Cutting Forces
Each set of 3-D dynamic cutting force was used to develop A R V ( n ) models first.
ARV(9) models were found adequate for the data collected under cutting condition
115
Group 1 in Table 5.1 and ARV(ll) models for the data collected under cutting
condition Groups 2-4. Given below is an example of ARV(9) model.
Xt= .448-.063-.020 .140 .808 -.018 -.034 .128 .333
X,i+
-.129 .115-.108 -.063-.349-.062 -.043-.083 .265
.262 .013 .193 Xt_2+|-.169.092 .010
.164 -.060 .052 Xt-3+
342-.023-.073* 103 -.034-.021 .085 .105 .048
Xt4f 187-.001 .005" .162.188 .088 .061-.054 .047
x** .018 .059 .099 .084-.058-.045 .004 .011 -.105
X«+
.200-.161 035 -.021 .108 .054
.010" -.016 .179
Xtf*
.086.158-.028" 054-.064-.035 086-.161-.131
Xt*
.084 -057 -.010
.003 .012 .105
.044 .060 -.003 Xt^+at
5.4.2 Patterns of Dispersion Development
After an adequate ARV(n) model was determined, dispersion tiYs and corresponding
frequencies were calculated according to Equations 5.4 to 5.8. The dominant
dispersions dj's, i.e., the ones with larger percentage Dj's, and the associated
frequencies under cutting condition Groups 1-3 are tabulated in Table 5.4. It is seen
that the most significant dispersions are associated with a lower frequency (LF) range,
and the second most significant dispersions related to a higher frequency (HF) range.
The dominant dispersions for all the first three cutting conditions are plotted in Figures
2.7 and 2.8, from which recognisable trends are observed which will be exploited in
the following section.
1 1 6
Table 5.4 Dispersion analysis results (only dominant terms listed)
Dispersion Percentages
Time
(minute)
0
2.67
5
10
15
20
25
34
Feed Force Fx
500-550 3.4-3.5
(HZ) (KHZ)
98.28
89.37
72.77
63.96
65.14
82.90
87.74
60.00
1.35
7.28
16.79
38.25
11.68
9.21
8.80
20.83
Dispersion Percentages
Time
(minute)
0
2.58
5.08
7.5
10
12.42
15.16
Feed Force Fx
450-550 3.3-3.5
(HZ) (KHZ)
86.06
66.5
61.3
50.50
35.50
65.5
75.20
8.57
10.20
21.35
20.36
24.20
15.57
6.72
Dispersion Percentages
Time
(minute)
0
2.58
5
7.5
10
12.67
15
Feed Force Fx
400-500 3.3-3.5
(HZ) (KHZ)
81.82
70.47
66.17
66.93
68.10
84.17
96.34
8.68
15.58
18.13
24.67
17.32
1.60
0.82
Di (%) for Cutting
Thrust Force Fy
650-750 3.3-3.5
(HZ) (KHZ)
55.43
44.66
61.60
79.84
76.43
72.39
61.98
74.32
12.11
13.42
14.77
16.22
20.80
21.90
31.74
20.32
Di (%) for Cutting
Thrust Force Fy
600-700 3.3-3.5
(HZ) (KHZ)
75.98
82.90
77.06
67.92
44.15
49.81
56.43
11.24
12.44
11.80
20.96
44.24
46.05
30.14
Di (%) for Cutting
Thrust Force Fy
550-650 3.3-3.5
(HZ) (KHZ)
78.71
81.99
76.05
61.54
46.10
47.14
38.68
12.99
11.05
14.97
28.60
37.97
42.51
12.96
Condition Group 1
Main Cutting Force Fz
950-1050 2.6-2.8
(HZ) (KHZ)
50.20
46.22
65.4
70.70
77.49
85.85
80.00
59.17
0.80
2.53
6.50
11.23
8.53
2.70
0.01
5.36
Condition Group 2
Main Cutting Force Fz
900-1200 3.0-3.5 i
(HZ) (KHZ)
27.70
39.53
57.38
24.15
40.54
66.28
81.53
1.59
1.75
7.78
2.54
19.64
16.61
1.03
Condition Group 3
Main Cutting Force Fz
800-1200 3.0-3.5
(HZ) (KHZ)
28.00
39.56
48.36
48.32
57.94
61.2
90.66
0.68
8.28
7.09
15.33
21.93
7.82 1 3.31
117
100 v x LL
C O
Cfl fc.
a> Q. CO
10 15 20 25 30
(a) LF=500-550Hz, HF=3.4-3.5KHz
o LF • HF
35
100
5 10 15 20 25 30
(b) LF=650-750Hz, HF=3.3-3.5KHz
o LF • HF
35
N U.
c o 0)
a> a
5 10 15 20 25 30
(c) LF=950-1050Hz, HF=2.6-2.8KHz
Time (minutes)
o LF © HF
35
Figure 5.7 Dispersion diagram for cutting condition Group 1
118
Cfl t_
0) Q. CO
100
80
60
40 "
20 -
L
-
• •
" ^ ^ d * •
I
LF HF
• **»
•
n /
•.^Sv»
• \
•
•
-
i--
•
**%
D
__!_
a LF • HF
a
a
-""S""""""
•
us
0 5 10 1 5 0 5 10 15 (a) LF=400-500Hz, HF=3.3KHz (a) LF=450-550Hz, HF=3.3-3.5KHz
1 ^^^
•
1 — d * - * ^ • 1
•
.
• •
1
LF HF
•
•
3
0 5 10 15 0 5 10 15 (b) LF=550-650Hz, HF=3.3-3.5KHz (b) LF=600-700Hz, HF=3.3-3.5KHz
100
80
60 h
40
20
0 0 5 10 15
(c) LF=800-1200Hz, HF=3-3.5KHz
Time (minutes)
— Group 2 —
-
•
D •
-
JT^*r
II
• - * * • * — " •
LF HF
n • a
•
D
0 5 10 15 (c) LF=900-1200Hz, HF=3-3.5KHz
Time (minutes)
— Group 3 —
Figure 5.8 Dispersion diagrams for cutting condition Groups 2 & 3
1 1 9
5.4.3 Analysis Associated with Physical Interpretation
Feed Direction: For an oblique turning operation of a bar, it is known that the feed
force F x is primarily associated with the normal force acting on the major flank F ^
and the horizontal friction force acting on the minor flank Fph (Figure 5.9). Therefore,
the tool/workpiece interactions on both flanks should be reflected in the dynamic feed
force characteristics.
Figure 5.9 The model for the forces acting on different tool faces
By examining the trend of L F dispersions of 500-550Hz for cutting condition Group
1 shown in Figure 5.7(a), it is found that the percentage values decrease to a minimum
between 10 to 15 minutes (major flank wear V B = 0.35 m m , Figure 5.4(a)), after
which they increase. It is well known from experience that cutting tools are replaced
or changed when the major flank wear reaches the critical values of 0.25-0.38 m m
[10, 86]. Beyond this critical wear, the rate of wear increases very rapidly, below it
the rate first decreases and then becomes constant. Thus, the behaviour of the L F
dispersions isolated from the dynamic feed force is very similar to the well-known rate
120
of major flank wear curves and could be used as a good indicator for major flank
wear.
By comparing the HF dispersion curve of 3.4-3.5 KHz shown in Figure 5.7(a) with
the minor flank wear VB' curve shown in Figure 5.4(c), it is again found the former
resembles the slope (rate) of the latter. The acceleration of VB' at about 10 minutes
could be detected by the maximum value of the H F dispersions.
Thrust Direction: The dynamic thrust force Fy mainly reflects the tool/workpiece
interactions on both the rake face and the minor flank. The small depth of cut used in
finish-machining produces a large chip flow angle such that the rake face friction force
F^ is almost along the y direction. The normal force acting on the minor flank Fpn is
also associated with Fy. In a similar manner, the H F dispersions of 3.3-3.5 K H z
shown in Figure 5.7(b) can be related to the rate of crater wear K T shown in Figure
5.4(b), and the L F dispersions of 650-750 H z shown in Figure 5.7(b) related to the
rate of the minor flank wear VB' shown in Figure 5.4(c). Therefore, they can be used
for minor flank and crater wear monitoring purposes.
Cutting Direction: The dynamic cutting force F z is primarily associated with the
normal force acting on the rake face F^, the friction force acting on the major flank
Fa, and the vertical friction force on the minor flank F«v. By examining the L F and
H F dispersions of 950-1,050 H z and 2.6-2.8 K H z shown in Figure 5.7(c), it was
found that they reflect the rate of the crater wear K T and the minor flank wear VB'
shown in Figures 5.4(b) and 5.4(c), respectively.
As summarised in Table 5.5, the trends of the LF dispersions isolated form all three
components of the dynamic cutting force reflect the wear rate mechanism associated
with normal forces, and the H F dispersions reflect the wear rate mechanism associated
with tangential (friction) forces. Similar results were obtained for experiments under
cutting condition Groups 2 and 3 as shown in Figures 5.5 and 5.8.
121
Table 5.5 Tool wear analysis for cutting condition Group 1
Fx
Fy
Fz
F a n (Normal to Major Flank) <->
Fgh (Tangential to Minor Flank) <-»
FRn (Normal to Minor Flank) <-»
F Y (Tangential to Crater Face) <->
Fy„ (Normal to Crater Face) <->
F R V (Tangential to Minor Rank) <->
F a (Tangential to Major Flank)
VB.KS
VB',N
VB'.N
KT
KT
V B \ N
<-»
<-»
<->
<->
<->
<->
Noi
LF Dispersions (500-550 Hz)
H F Dispersions (3.4-3.5 KHz)
LF Dispersions (650-750 Hz)
H F Dispersions (3.3-3.5 KHz)
LF Dispersions (950-1050 Hz)
H F Dispersions (2.6-2.8 KHz)
recognisable trend was found.
5.4.4 Strategies of Tool W e a r Estimation in Finish-Machining
Major Flank Wear: Based on the discussion above, it can be concluded that the LF
dispersion of feed force F x is in agreement with the rate patterns of major flank wear
V B , thus giving a good indication for the major flank wear.
Minor Flank Wear : Among the three dispersion patterns for minor flank wear, i.e.
the L F dispersion of Fy, the H F dispersion of F x and the H F dispersion of Fz, the
last two are found to exhibit a consistent pattern under all the three cutting conditions,
Groups 1-3. The H F dispersion of Fx, however, is more a static than a dynamic one,
because of the slow feed motion. Therefore, H F dispersion of Fz can be used as the
main indicator of minor flank wear VB', and the H F dispersion of Fx as auxiliary
one. W h e n one of them reaches the maximum, the accelerated minor flank wear is
indicated.
Crater Wear : For the crater depth, it can be seen that both LF dispersion of Fz and
H F dispersion of Fy under all the three cutting conditions, Groups 1-3, reach their
maximum values as the crater depth K T , falls into the accelerating stage. Thus both
of them could be used as an indicator for the detection of the accelerating state of crater
depth.
122
5.4.5 Structural Dynamics and Idle Disturbances
Clear patterns linking the force variations in terms of dispersions and associated
frequencies, isolated from dynamic cutting force, to the various wear development
rates have been identified. However, the physical nature of the relationships is
unclear and it is the purpose of this section to identify physical origins of these
relationships and interpret accordingly.
Since almost the same HFs appear in all three groups, these frequencies are then
inherent in the tool holder and hence may be conjectured to relate to its natural
frequencies. Tests revealed that the natural frequencies of the tool
holder/dynamometer system were 3,320 Hz in the x-, 3,300 Hz in the y-, and 2,847
Hz in the z- directions, respectively. These values match reasonably well with the
HFs isolated from the dynamic cutting force (Table 5.4).
Tests on dynamometer frequency response to idle speed excitation alone revealed idle
frequencies of 575 Hz for x, 715 Hz for y, and 975 Hz for z under cutting condition
Group 1. These frequencies are reasonably close to the LF's listed in Table 5.4.
From the above tests, it becomes clear that tool/workpiece interaction at the LFs are
related to the idle frequencies, and the HFs are mainly associated with the natural
frequencies of the tool-holder/dynamometer system.
5.5 SUMMARY AND CONCLUDING REMARKS
(1) Dispersion analysis based on trivariate ARMAV time series models was used to
decompose quantitatively the dynamic cutting force in terms of dispersions
(relative importance of modes of force variation) and the associated frequencies.
The merit of the method is its ability to isolate, from the dynamic cutting force,
the ingredients, each of which is sensitive to a particular wear state, thereby
123
providing much more comprehensive yet sensitive estimates than those possible
by using the force signal in a lump-sum manner.
(2) The patterns of change of the dispersions resemble the rate of various wear
parameters and the resemblance is physically interpreted. The rapidly increasing
rate of minor flank wear occurring before the accelerating stage of major flank
wear is due to the fact that the gradually increasing major flank wear, and retreat
of the cutting edge, sharpens the nose and puts more burden on the minor flank
and edge, to a point where drastic minor flank wear is inevitable. Therefore, for
operations such as a finish-turning of a bar, optimum cutting conditions or
effective tool change strategies have to be determined based on the minor flank
wear instead of others, to assure geometric accuracy and surface quality of the
finished workpiece.
(3) From a practical machining viewpoint, the algorithm introduced is feasible for
on-line tool wear estimation after reformulation and simplification because it is
sufficiently fast to determine tool wear states in real time.
(4) Different machine tools or dynamometers certainly have different frequency
characteristics which, however, can be determined with reasonable ease. The
trends identified by using dispersion analysis in this chapter remain the same
while the specific values of these structurally-determined frequencies may vary.
In this sense, the approach is applicable to the general case of finish-machining.
124
CHAPTER 6
MONITORING GROOVE WEAR DEVELOPMENT IN FINISH-MACHINING BY ARV MODEL-BASED
MULTIPLE DISPERSION ANALYSIS
6.1 INTRODUCTION
As described in Chapter 5 for finish-machining the well-known types of tool wear,
such as major flank wear and crater wear, are often not the wear types which lead to
tool failure first. Instead, minor flank wear, nose wear and groove wear at the minor
cutting edge are recognised as being more important in determining the tool life, due to
their greater influence on dimensional accuracy and surface quality of the finished
product [87], Particularly, groove wear at the minor cutting edge, once formed, will
cause significant deterioration of surface quality and shorten tool-life £23-251.
Therefore, it is obvious that for finish-machining, monitoring of the minor flank,
major flank and crater wear is not sufficient and monitoring of the groove wear at the
minor cutting edge must be incorporated. Since the groove wear often induces high
frequency vibration, 3-D vibration signal is chosen for monitoring purpose.
The study of groove wear at the minor cutting edge can be dated back to 1960's [23-
26, 90-92]. Most work then focused on explaining the phenomena and conjecturing
the mechanism of formation. The formation of the groove is mainly due to the rubbing
action at the minor cutting edge from a mechanical point of view. The workpiece, with
a work-hardened layer, can be compared to a grinding wheel. Groove wear will occur
under certain combinations of tool and work materials, and it can be observed that a
series of grooves will form at the minor cutting edge which in fact does not take part in
cutting directly. The depth of groove wear directly influences the surface roughness
produced. Shown in Figure 6.1 is a pictorial description of the groove wear at the
125
minor cutting edge in accordance with CIRP terminology [87]. It was found that when
enough grooves have been formed and developed to a certain degree, severe vibration
may be induced, which disturbs the formed groove pattern and wipes out the grooves
eventually, resulting in a rapid deterioration of surface finish [26]. In other words,
groove wear reaches its critical point and the cutting tool should be replaced when the
grooves are being wiped out.
(a) View from Tool Top Face (b) View from A
Figure 6.1 Groove wear at the minor cutting edge
Although vibrations induced by groove wear at the minor cutting edge have been
observed long ago [23], no further work has been reported on investigating the
possibility of using it for on-line monitoring purposes, mainly due to the complexity
involved in groove wear formation and the difficulty involved in effective signal
analysis and interpretation. The study of on-line groove wear monitoring, however, is
of significance in automated machining systems, not just because of its influence on
surface quality, but also because of the fact that groove wear is often one of the factors
inducing unfavourable chatter [25]. Although it was reported that superimposing a
minor vibration in the feed direction hindered groove formation {25-26], this is not
economically justifiable for conventional machine-tool structures.
1 2 6
The purpose of this chapter is to develop an effective method to detect and monitor the
formation and development of groove wear. The experimental results of groove wear
development and investigation results concerning the detection of the critical point at
which severe vibration is induced by the groove wear, i.e. at which the finishing tool
needs to be replaced, are presented first. Due to the fact that the vibration occurring in
the neighbourhood of that point in time is extremely complicated and that very little is
known about it, a miniature 3-D accelerometer, mounted in close vicinity to the tool
tip, was used to capture multi-dimensional vibration signals. Since more than one
quantity is needed to describe the groove wear, and the formation of groove wear is
related to all three orthogonal cutting directions, multivariate autoregressive time series
models (ARV) are adopted. Based on the stochastic A R V models developed directly
from the vibration signals, a quantitative analysis was made possible by employing
multiple dispersion analysis, which discriminates features which are sensitive to
various aspects of the formation of groove wear.
6.2 INVESTIGATION INTO THE PATTERNS OF GROOVE WEAR DEVELOPMENT
6.2.1 Tool Wear Experiments
In the experiments 3-D vibration signals were measured by a miniature 3-D
accelerometer (PCB Model 306A06), mounted in close vicinity of the tool tip, aiming
at capturing original signals with minimum distortion. The machining conditions used
in the experiments are shown in Table 6.1, all the conditions being within the range
recommended by the tool manufacturer.
127
Table 6.1 Machining conditions used in groove wear experiments
Machine Tool
Tool Insert Type
Tool Geometry
Work Material
HITEC-20Sn C N C Lathe (18 K W ) j
TNMG160408
(Carbide P10 and groove-type chip breakers)
0°, 5°, -6°, 90°, 60°, 0.8
AISI4140 (BHN=320)
Cutting Conditions :
Group A : V=160rn/min f=0.08 mm/rev d=0.25mm
Group B : V=160rn/min f=0.04 mm/rev d=0.25mm
Group C : V=125m/min f=0.08 mm/rev d=0.25mm i
Group D : V=190m/min f=0.08 mm/rev d=0.25mm
Group E : V=190m/min f=0.04 mm/rev d=0.25mm
The machining process was interrupted periodically in order to measure tool wear and
surface roughness. T w o sets of 524 data points each, one with a sample interval of
30ps and the other with 3.13ms, were taken from the vibration signal in each of the
three orthogonal directions, just before each interruption. A n extra two sets of data
were also recorded between consecutive interruptions to provide more information for
signal processing.
6.2.2 Development of Groove Wear
In order to describe quantitatively the development of groove wear at the minor cutting
edge four parameters were selected, viz. the number of grooves at the minor cutting
edge, the depth of grooves, the maximum length of grooves, and the groove wear area
(defined as the product of the maximum groove length and the distance between the
first and the last grooves on the minor flank). Nose wear was also selected due to its
relevance to groove wear. Figures 6.2 to 6.6 show the measurement results of these
wear parameters under all five cutting conditions. The surface roughness was plotted
in Figure 6.7. All these data are listed in Appendix C for details.
8 -
6-
4-
2 -i
Oi f—i 1- r — i - i __ r _ r r - r "'—i
Group A
- Group B
- Group C
Group D
Group E
0 5 10 15 20 25 30 35 40 45 50 55 60 Time (min)
Figure 6.2 Number of grooves at the minor cutting edge
12
10-
8-
Q. © c Q 6 CD > O O A •- 4
0 2-
o Group A
• Group B
• Group C
x Group D
A Group E
— i — i — i — i — i — i — i — i — i — i — — i — i — i — r — i — i —'—r—' 10 15 20 25 30 35 40 45 50 55 60 Time (min)
Figure 6.3 Development of groove depth
400
T — ' — i — ' — i — « — i — ' — i — ' — i — • — i — ' — i — • — i — • — i — i — i — • — r 0 5 10 15 20 25 30 35 40 45 50 55 60
Time (min)
Figure 6.4 Development of the maximum groove length
i—'—i—i—i—i—i—"—i—i—i—"—i—•—i—«—i—"—i—i—i—i—r 5 10 15 20 25 30 35 40 45 50 55 60
Time (min)
Figure 6.5 Development of groove wear area
- I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I -
0 5 10 15 20 25 30 35 40 45 50 55 60 Time (min)
Figure 6.6 Development of nose wear
E ^3 CC co co CD
c -C § 2 cr CD O
•g 3 w
o Group A
. • Group B
• Group C
- x Group D
A Group E
X
TA
• • # • °k ° ° * A „ ox •
A
O X X
MM1^1 • • • •
. •
— ' — I - 1
10 15 20 25 30 35 40 Time (min)
45 50 55 60
Figure 6.7 Surface roughness for five cutting conditions
131
6.2.3 Patterns of Groove W e a r Development
From an analysis of the above figures, it is seen that two different groove wear
patterns can be identified, one for the lower feed (0.04 mm/rev), designated as Groups
B & E, and the other for higher feed (0.08 mm/rev), designated as Groups A, C & D.
For the lower feed, it can be seen the number of grooves decreases at a certain point
(Figure 6.2) and the surface finish deteriorates sharply at the same time (Figure 6.7).
It is evident that the wipe-out of grooves causes the deterioration. It is also seen that
the depths of grooves fluctuates for the low-feed Group B (Figure 6.3) because of the
concurrence of the formation of new grooves and the wipe-out of formed grooves,
while for the higher feed no decrease in the number of grooves nor a sharp increase in
surface roughness are observed, even after machining for over 45 minutes (Figure
6.2). This may be due to the fact that the higher feed rate results in wider ridges
between adjacent grooves, and thus in better wear-resistance.
It is consistently observed for all five cutting conditions that when the groove wear at
the minor cutting edge has developed to a point at which tool failure is indicated, the
major flank wear is still far below its critical value. Therefore, it can be concluded that
the groove wear, once formed at the minor cutting edge, will largely dictate the tool life
in finish-machining. A scanning electron microscope (SEM) was used for a more
detailed analysis, with some results being shown in Figures 6.8 and 6.9.
132
Group B (V=160 m/min) 50ium lOO^m
Group E (V=190 m/min)
o CO
co
ro
ra
©
ra
CO
>. T3 ra © CO
o O) ra CO 0) t_
> O CO
©
o ra
w O) c ra
o Q. Q. CO </>
Figure 6.8 Groove wear development under the lower feed condition (0.04 mm/rev)
133
Group A (V=160 m/min) 50^m 100jim
I 1 Group D (WI90 m/min)
©
CO
CO
co
co
©
a CO
>>
CO ©
CO
© CB a CO CO
c © ©
u o <
Figure 6.9 Groove wear development under the higher feed condition (0.08 mm/rev)
134
Lower feed (Groups B & E^
Four typical stages are found under the lower feed condition, as shown in Figure 6.8.
(a) Initial stage — Before the grooves are formed, the surface roughness is mainly
determined by the tool geometry of the fresh tool as well as the cutting conditions
used, usually representing the best possible surface roughness condition in finish-
machining.
(b) Steady stage ~ In the next few minutes, the grooves will form at the minor
cutting edge under certain machining conditions. Once these grooves are formed, the
surface roughness will increase and largely depend on the depth of the grooves. As
the depth of grooves develops slowly during this period, the surface roughness
remains constant
(c) Severe stage — From Figure 6.8(c), it is clear that at this stage the grooves are
being wiped out due to the induced vibration. W h e n the regular groove pattern is
disturbed, the surface roughness increases sharply, resulting in a very poor surface
quality.
(d) Disappearing stage — At this stage, most grooves have been wiped out and the
minor flank face is left in a very rough condition and loses its function as a finishing
tool.
Higher feed (Groups A. C & D^
For the higher feed, no decrease in the number of grooves occurs and three typical
stages as shown in Figure 6.9 may be identified.
135
(a) Initial stage - At this stage, groove wear development is similar to that under the
lower feed, while the surface finish is poorer due to the higher feed used.
(b) Steady stage - There seems to be no significant difference in surface roughness
compared with the lower feed group. However, the grooves formed at the minor
cutting edge are fewer due to the higher feed used.
(c) Accelerating stage - Shown in Figure 6.9(c) is the accelerating stage of groove
wear, during which the surface roughness (Figure 6.7) begins to increase quickly and
soon becomes unsuitable for finish-machining. Viewed together with Figure 6.5,
which shows the development of groove wear area, it is found that the starting point of
rapid deterioration of surface finish always corresponds to the beginning of
acceleration of the groove wear area, indicated by the dotted lines in Figure 6.5.
Therefore, it is conjectured that the rapid increase of surface roughness at the high feed
is due to the larger contact area between the tool minor flank and the machined
workpiece surface.
6.3 ARV Model-based Multiple Dispersion Analysis
6.3.1 Application of Autoregressive Vector Time Series Modelling
Autoregressive vector time series models were developed to quantify statistically the
dynamics embedded in the 3-D vibration data. It has been shown in Chapter 5 that the
merit of employing such a technique lies in the fact that it provides a mathematical
model primarily based on the observed data, and does not require much a prior
knowledge or assumptions about the underlying system dynamics. The vibration
136
signals, recorded in three orthogonal directions and sampled at uniform intervals, can
be represented in the form of vector difference equations, i.e. either in an explicit
Green's function or in a stochastic ARV(n) model with autoregressive order n [35]:
n oo
Xt=X<PkXt_k + at or Xt=£Gkat_k (6.1) k=l k=o
r T T where X t = [ X l t , X 2 t , X 3 t ] and at = [ al t , a2t, a3t] .
In this way, the observed trivariate series, Xlt = Vxt = vibration in the feed direction,
x2t = vy t = vibration in the thrust direction, and X 3 t = Vz t = vibration in the main
cutting direction, are expressed either as linear combinations of past observation
vectors Xt.k with parameter matrices Ok's (k=l, 2,..., n) plus the white noise at, or
in terms of the Green's function G k by convolution with the white noise vectors at.k
[35, 93].
6.3.2 Multiple Dispersion Analysis
In Chapter 5 the use of dispersion analysis provides an effective means for quantitative
model decomposition in order to identify particular variations convolved in the
recorded data of 3-D dynamic cutting force. Since the formation of groove wear may
stem from complex interactions among all three dimensional vibrations and much about
it is still not clear, multiple dispersion analysis is used to quantify not only the
137
contribution of individual variables, but also the interactions among different variables.
Using the explicit Green's function of Equation 6.1, the vector auto-covariance matrix
y0.vec can be determined as follows :
Yo-vec=E[XfXtT]= lGk.a'.Gk
T (6.2) k=0
where aa2 is the residual matrix and Gk is the complex conjugate of Gk. It has been
shown that the explicit Green's function can be obtained from the established ARV(n)
model under the assumption of distinct eigenvalues [93] as follows,
JL, k+n-l , ,
Gk = £Ti^i U^UiTi (6.3) i=l
n n
where U= n ( W *"* Ui=("1)l H ( W i, j=l & i>j i,j=l & i>j & i. j k
The eigenvalue matrices Xi = diag[X,xi, X,yi, Xz(\ and eigenvector matrix Ti, i=l,...,
n, are found by adjoining the parameter matrices Ok's of the ARV(n) model, and
finally the following form of the equation for a symmetric vector auto-covariance
matrix Y0_Vec can be derived:
138
• o-vec
•oxx »oxy 11 y «oxz
oyy «oyz
• ozz
IOVV l|
n n n n n n IX D ixxj UPixyj 22>ix«j i=lj=i i=lj=l i=lj=l
n n n n Z^dZA^iyyj dZjdZJ^iyz] i=lj=i i=lj=l
n n
IIP i=lj=i
1ZZJ
(6.4)
Thus the vector auto-covariance of the entire process is decomposed into the
contributions of process eigenvalues as well as the complex interactions among various
eigenvalues in terms of multiple dispersion D ^ , Dixyj, etc. Of particular interest are
the multiple dispersions corresponding to the complex eigenvalues which reflect the
oscillating characteristics of the machining process.
6.4 STRATEGY FOR GROOVE WEAR MONITORING
BY MULTIPLE DISPERSION ANALYSIS
The 3-D vibration signals were modelled as a stochastic A R V ( n ) model. The
multivariate time series analysis and the results of multiple dispersion analysis are
presented below in relation to the detection and monitoring of groove wear
development at the minor cutting edge.
6.4.1 Groove Wear Monitoring for the Lower Feed Condition
Three distinctive dispersions were found as being related to the development of groove
wear, as shown in Figure 6.10 for Group B as a representative of the lower feed
condition (0.04 mm/rev). The first two patterns are associated with high-frequency
dispersions. One is the dispersion in the thrust direction (Dyy) at around 9.3 K H z and
139
another is that in the main cutting direction (Dzz) at around 2.5 K H z . It was found that
when the number of grooves decreased from about 16 to 23 minutes (Figure 6.2), Dyy
and Dzz reached their maximum values, signifying the severe vibration occurring under
these frequencies. The third pattern is the cross-dispersion D y z in the low-frequency
range between the vibrations in the thrust direction at 150Hz and the main cutting
direction at 145Hz (Figure 6.10). The moment at which D y z reaches its peak value
also matches with that when the number of grooves decreases and the roughness
deteriorates rapidly. If one compares the times of peak dispersions, it is seen that D j ^
occurred at about 19 minutes and Dzz and D y z at about 22 minutes. This may indicate
that vibration is first induced in the thrust direction which disturbs the regular grooves
formed. The disturbed minor cutting edge in turn excites vibration in the main cutting
direction. Therefore, the following criterion can be established: the peak value of D y y
indicates the commencement of groove wipe-out, while those of D ^ and D y z indicates
the ending of groove wipe-out. This criterion can be used in finish-machining to
determine the necessity of tool replacement for the lower feed condition.
100-1 1 1
Time (min)
Figure 6.10 Dispersion patterns for the lower feed condition (0.04 mm/rev)
140
6.4.2 Groove Wear Monitoring for the Higher Feed Condition
For this condition, the dispersion pattern found is totally different from that for the
lower feed, due to the difference in groove wear development. A clear concave trend
shown in Figure 6.11 was found for the dispersion in the thrust direction, with the
frequency equal to about 9.4 KHz. B y comparing the surface roughness development
(Figure 6.7) with the above dispersion pattern, it is seen that consistency exists
between them, as summarised in Table 6.2.
100
80
• Group A n Group C x Group D
0 5 10 15 20 25 30 35 40 45 50 55 60 Time (min)
Figure 6.11 Dispersion patterns for the higher feed condition (0.08 mm/rev)
Based on the analysis in Table 6.2, it can be concluded that the behaviour of the
dispersion in the thrust direction (9.4 KHz), isolated from the 3-D vibration variations,
resembles the rate (slope) of the surface roughness development shown in Figure 6.7.
Thus this dispersion could be used as an index for the severity of groove wear at the
minor cutting edge under higher feed conditions.
1 4 1
Table 6.2 Relationship between surface roughness and dispersion pattern
Groove
Wear
Development
Initial
Stage
Steady
Stage
Accelerating
Stage
Changes of Surface
Roughness for the
Higher Feed Group
increasing rapidly due to
the normal tool running-in
remaining constant
as the groove depth
develops slowly
starring to accelerate due to
the severity of groove wear
indicated by the acceleration
of groove wear area (Fig. 6.5)
Dispersion Pattern around
the Natural Frequency of
Tool/Holder Assembly
appearing in high values
during tool running-in period
decreasing to the minimum
due to smaller variations
involved in the thrust
vibration signals
increasing to the high-value
again due to the changes
in thrust vibration caused
by the severe groove wear
6.5 PHYSICAL INTERPRETATIONS
Clear patterns have been identified relating the vibration in terms of multiple
dispersions and corresponding frequencies to the different stages of groove wear
development. For the lower feed condition (0.04 mm/rev), more grooves are formed
and earlier vibration is induced, giving rise to worse surface finish, while the higher
feed condition (0.08 mm/rev) appears to have better wear-resistance. The difference
between the lower and higher feed conditions lies in the fact that for a selected tool
geometry, the feed becomes a dominant factor to determine the grooves' spacing at the
minor cutting edge. It is noticed that, in the case where grooves are formed, the lower
feed condition does not produce better surface finish.
142
For both conditions the dispersion analysis results, based on the stochastic model
developed from the experimental data directly, indicate that groove wear development
finds its reflection mainly in the thrust vibration. This can be explained as follows.
This vibration, i.e. the relative displacement between the workpiece and the tool tip in
the radial direction, is a compound effect of the workpiece lateral dynamics and the
tool/tool holder dynamics. In the analysis for the lower feed condition it was shown
that the dispersions of the thrust vibration reached their peaks at about 150 H z and
9,300 H z respectively, with the higher one dominant when the number of grooves
began to decrease, that is, the surface roughness began to deteriorate rapidly. These
frequencies, as expected, correspond to the natural frequencies of the workpiece and
tool/holder assembly, which were determined to be around 155 H z and 9,340 Hz,
respectively, by using conventional excitation tests. These close agreements may
indicate that, when a sufficient number of grooves have been formed, the dynamics of
the cutting process is excited, so that more severe vibration at the characteristic
frequencies is induced.
Similar results were obtained in the dispersion analysis of vibration in the main cutting
direction. For the high feed condition, the dispersions in the thrust direction around
the natural frequency of tool/holder assembly were again found closely related to the
surface roughness development determined by the severity of groove wear at the minor
cutting edge. Table 6.3 summarises these physical interrelationships.
143
Table 6.3 Comparison between dispersion analysis results and physical quantities
For the Lower Feed Condition (0.04 mm/rev)
Vibration
Direction Thrust
Main Cutting
Peak Frequencies detected
by using Dispersion Analysis
when the number of grooves
began to decrease
in the range of
low frequency 150 H z
145 H z
in the range of
high frequency 9,300 H z
2,500 H z
Natural Frequencies
determined by using
conventional excitation !
tests
workpiece
system 155 H z
155 H z
tool/holder
assembly 9,340 H z
2,610 H z
For the Higher Feed Condition (0.08 mm/rev)
Thrust Vibration
Characteristic Frequency of
Dispersion which resembles
the rate of Surface Roughness 9,400 H z
Natural Frequency
of the tool/holder
assembly 9,340 H z
6.6 SUMMARY AND CONCLUDING REMARKS
(1) The purpose of this chapter is to investigate the relationships between off-line
measurements of groove wear at the minor cutting edge and the multiple
dispersion analysis based on the multivariate time series models of 3-D vibration
in finish-machining, such that the latter alone will be capable of on-line monitoring
of groove wear development at the minor cutting edge.
(2) The development patterns of groove wear at the minor cutting edge and the
associated surface roughness were investigated under typical cutting conditions.
It was found that if a finish-machining condition under which grooves will form at
144
the minor cutting edge is used, the feed is the dominant factor in deterrnining the
development patterns of groove wear.
(3) Multiple dispersion analysis of the 3-D vibration has been proven to be effective in
detecting the critical points at which either the surface finish deteriorates sharply
due to the wiping-out of grooves at the minor cutting edge, or the acceleration of
surface roughness induced by the rapid increase of groove wear area occurs. The
criteria derived from the dispersion analysis are all based on the frequencies which
can be physically interpreted. Although these frequencies may vary for different
workpiece geometries, and tool/tool holder assemblies, they are not difficult to
determine.
(4) The results of this work further emphasise that groove wear at the minor cutting
edge m a y occur during finish-machining under a fairly wide range of cutting
conditions for the commonly-used cutting tools and work materials. Should
grooves be formed, the life of the finishing tool will be significantly shortened.
(5) The monitoring method presented in this chapter offers one avenue by which
surface finish may be controlled in automated finish-machining, as an essential
component of product quality assurance.
145
CHAPTER 7
INTEGRATING CHIP CONTROL AND TOOL WEAR ESTIMATION THROUGH NEURAL NETWORKS FOR THE ON-LINE ASSESSMENT OF MACHINING PERFORMANCE
7.1 INTRODUCTION
Chip control and tool wear estimation are closely interrelated during machining
processes. Although much work has been done on the analysis and prediction of chip
forming patterns including chip shapes and chip breaking in machining [2,40, 50,94-
95], all assumed ideal machining conditions, i.e. machining with unworn cutting
tools. It is also known that present theories concerning chip formation in machining
and available machinability databases are all established based on unworn tools. In
actual machining processes, however, the chip forming patterns vary significantly
with tool wear progression, thus resulting in unpredictable performance of the
machining operation. In this sense, a truly effective chip control system should not
only be capable of predicting chip forming patterns off-line, but also updating them
on-line as tool wear develops in the machining process.
During the machining process tool wear formed at different tool faces alters the
original tool configuration/geometry, which, in turn, greatly influences chip forming
patterns. Therefore, in order to assess chip forming patterns with wear progression,
an effective estimation strategy for tool wear at different tool faces is a prerequisite. In
Chapter 5, dispersion analysis of 3-D dynamic cutting force derived from multivariate
time series models has proven particularly effective for comprehensive tool wear
estimation, i.e. more than one type of wear can be estimated simultaneously, thus
146
providing a possible basis for predicting chip forming patterns during tool wear
progression.
Since the interrelationship between chip forming patterns and tool wear progression is
extremely complex and the present machining theories are inadequate to describe it
analytically, some kind of "black-box" approach becomes appealing, even necessary,
for tackling the problem.
One such approach, neural networks, has been applied recently in machining process
monitoring or control [11-12, 37-38]. Rangwala and Dornfeld [11] demonstrated the
feasibility of using neural networks to integrate information from acoustic emission
and force sensors to monitor flank wear, where the features from power spectra were
extracted as the inputs to neural networks. In subsequent work, Dornfeld [12]
developed a neural network-based tool wear monitoring system by incorporating
autoregressive time series model parameters. Chryssolouris and Domroese [37]
performed a simulation system to evaluate the learning ability of neural networks and
highlighted the use of neural networks as the decision-making component in an
intelligent tool wear monitoring system. Rangwala and Dornfeld [38] presented a
method for optimising machining conditions based on the neural network
architecture, which has shown that the neural network can learn and synthesise
knowledge effectively by observing the input variables of cutting conditions and the
output variables of cutting force, power, temperature and surface finish.
Therefore, neural networks provide a new approach to resolve a complicated problem
through learning by being shown and synthesising knowledge from the observed
input and output variables of the process under investigation.
147
For the chip forming pattern problem, chip shapes produced are observable and
readily collectable, while tool wear can be estimated with higher certainties by the
monitoring strategy developed in Chapter 5. Therefore, employing neural network
techniques, which model a complicated process by extracting knowledge largely based
on experimental input-output data, would be feasible to correlate dynamic chip
forming patterns with tool wear progression. In addition, once it is trained off-line,
the established network algorithm can be implemented on-line.
In Chapter 2, it has been shown that prediction of chip breaking/shapes under the
condition of unworn cutting tools can be achieved through a fuzzy membership
function ranging from 0 to 1, which provides the initial estimation of chip forming
patterns. Neural networks are then trained based on the recorded data of chip
breakability/surface roughness in finish-machining with tool wear progression. The
latter is characterised by four features extracted from multivariate time series models of
3-D dynamic cutting force signals. Using the results derived from neural network
modelling, together with the comprehensive tool wear estimation, an integrated
assessment of machining performance may be achieved, which includes chip breaking
and chip shapes, surface finish and tool wear states.
7.2 DESCRIPTION OF MACHINING EXPERIMENTS
7.2.1 Machining Conditions
A series of machining experiments concerning chip breaking and chip shapes, surface
finish and tool wear was carried out under finish-machining conditions to obtain the
data needed for training and testing the neural network to be established. Shown in
Table 7.1 are the machining conditions used in the experiments. The cutting
148
conditions are organised into two groups, i.e. training and testing. The degraded tool
tests, viz. using a relatively softer tool as recommended in [86], are adopted to shorten
the time-consuming and costly tool wear experiments.
Table 7.1 Machining conditions used in the experiments
Machine Tool
Tool Insert Type
Tool Material
Tool Geometry
Work Material
Depth of Cut
Colchester Mascot 1600 (9.3 K W )
TNMA160408F (flat-faced tool)
Carbide : Grade 883, S E C O
0°, 5°, -6°, 90°, 60°, 0.8
AISI4140 (BHN=275-320)
d = 0.5 m m
Cutting Speed and Feed
Training Group
1. V=l 15 m/min f=0.1 mm/rev
2. V=145 m/min f=0.1 mm/rev
3. V=145 m/min f=0.06 mm/rev
4. V=205 m/min f=0.06 mm/rev
5. V=170 m/min f=0.10 mm/rev
6. V=160 m/min f=0.15 mm/rev
Testing Group
1. V=140 m/min f=0.12 mm/rev
2. V=165 m/min f=0.06 mm/rev
3. V=190 m/min f=0.06 mm/rev
4. V = 130 m/min f=0.15 mm/rev
7.2.2 Comprehensive Tool Wear Patterns
Based on Chapter 5, patterns of major flank, crater and minor flank wear were
considered as they reflect the overall tool wear situation and largely influence tool
configuration/geometry and the surface quality of a finished product. Patterns of
major flank wear V B , crater wear (depth) K T and minor flank wear VB' are
summarised and typical results with Training Cutting Condition 1 as representatives
are plotted in Figure 7.1.
149
Time (min)
Figure 7.1 Development patterns of comprehensive tool wear
7.2.3 Chip Breakability Assessment
Chip breakability was assessed by using the fuzzy membership function according to
the chip shapes/sizes produced in the machining process. Figure 7.2 illustrates the
change of chip shapes/sizes with cutting time under Training Cutting Condition 1, and
Table 7.2 gives the description of these chips and the corresponding fuzzy
membership values assigned.
150
) J 1 /
i = ll minute
t = 15 minutes
0Z, /
\ t = 2.7 minutes
/WW
JVWW
A/WV
t = 20 minutes
^J^
t = 5 minutes
t = 25 minutes
t = 10 minutes
fVVVN/VA>L,
t = 34 minutes
Figure 7.2 Change of chip shapes/sizes with tool wear progression
Table 7.2 Chip forming behaviour in machining process while tool wear
TIME (min)
0
2.7
5
10
15
20
25
34
CHIP SHAPES/SIZES
long and continuous ribbon chips
curved ribbon chips
combination of ribbon and continuous cork-screw chips
long and continuous cork-screw chips
long but broken cork-screw chips
short to medium size chips
distorted medium size chips
CHIP BREAKABTLITY (MEMBERSHIP V A L U E )
0.25
0.30
0.32
0.35
0.41
0.58
0.46
heavily distorted long but broken chips 0.42
151
A s is seen from Fig 7.2 and Table 7.2, it is clear that tool wear has a significant effect
on chip forming behaviour. This is largely due to the fact that when tool wear
develops, tool configuration/geometry changes accordingly. In the initial stage of
machining, long and continuous chips form, because the tool insert used has no chip
breaker. With the increase of machining time, tool wear causes the change of tool
configuration/geometry. In particular, the formation of crater wear on the tool rake
face acts as a groove-type chip breaker and thus increases chip breakability. As crater
wear grows larger its effect as the chip breaker becomes more significant until a fully-
utilised chip breaker is realised which breaks chips in the most effective way as seen at
t = 20 minutes. Further growth of crater wear oversizes the chip breaker, resulting in
the increase of chip curling curvature which lowers chip breakability. The change of
chip breakability as tool wear develops has been observed being consistent under all
cutting conditions used, as shown in terms of fuzzy membership values in Figure 7.3
for Training Cutting Conditions 2 to 6.
1.0
e
.a E
0.8-
0.6-
• Condition 2: V=145m/min, f=0.1 Omm/rev - o — Condition 3 : V=145m/min, f=0.06mm/rev
- © — Condition 4 : V=205m/min, f=0.06mm/rev
• + — Condition 5: V=170m/min, f=0.10mm/rev
- n — Condition 6 : V=160m/min, f=0.15mm/rev
Time (min)
Figure 7.3 Change of chip breakability under Training Cutting Conditions 2-6
152
Although the above analysis clearly confirms the interrelationship between chip
forming patterns and tool wear progression, its analytical modelling proves
prohibitively difficult, even impossible. In order to predict the chip forming patterns,
a quantitative description is required, which highlights the need for introducing the
modelling by neural networks.
7.2.4 Surface Finish Assessment
Surface finish was assessed by the arithmetic mean deviation, Ra, with the use of a
portable surface measurement instrument. Although surface finish can be predicted
theoretically with some accuracy, its development depends heavily on the severity of
tool wear during the machining process, as shown in Figure 7.4 for Training Cutting
Condition 1. Compared with the tool wear development shown in Figure 7.1, it is
noticeable that the significant increase of surface roughness corresponds to the
acceleration stage of minor flank wear.
E
oc N
m m c •E 2-
o E • o a CO
initial accelerating stage of major flank wear VB
accelerating start of minor flank wear VB'
-r 5
accelerating start of crater wear KT and major flank wear VB
15 20 — f —
25 10 30 35 Time (min)
Figure 7.4 Change of surface finish with tool wear rates for Training Cutting Condition 1
153
Similar results were obtained under Training Cutting Conditions 2 to 6, as shown in
Figure 7.5.
0.5
• Condition 2 : V=145m/min, f=0.10mm/rev • Condition 3 : V=145m/min, f=0.06mm/rev o Condition 4 : V=205m/min, f=0.06mm/rev + Condition 5: V=170m/min, f=0.10mm/rev n Condition 6 : V=140m/min, f=0.15mm/rev
-r 5
—r~ 10
I
15 Time (min)
Figure 7.5 Development of surface finish for Training Cutting Conditions 2-6
7.3 FEATURE EXTRACTION FROM 3-D DYNAMIC CUTTING FORCES
In Chapter 5, dispersion analysis based on multivariate time series models has proven
to be effective to single out particular features in the 3-D dynamic cutting force signals
154
corresponding to particular types of tool wear. Four dispersion features extracted
from dynamic components of 3-D cutting forces are found to correlate reliably with
various types of tool wear development under all training and testing cutting conditions
(Table 7.1). The results are summarised below.
The eigenvalues appearing in complex conjugate pairs are of particular interest because
they contribute to the oscillating variation of the process. T w o dominant percentage
dispersions, one in low frequency (LF) related to the idle frequencies of machine-tool
and the other in high frequency (HF) related to the natural frequencies of the tool-
holder/dynamometer system, are found to exhibit patterns in agreement with wear rate
patterns including major flank wear V B , crater wear K T and minor flank wear VB'.
A s a summary, the relationship between wear rate patterns and dispersion
development patterns is shown in Figure 7.6.
TOOL WEAR TYPES major flank wear (VB)
crater wear (KT)
minor flank wear (VB')
TOOL WEAR RATE PATTERNS
Stage I
accelerating
steady
steady
Stage H
steady
accelerating
accelerating
Stage IB
accelerating
steady
steady
DESCRIPTIONS O F DISPERSION DEVELOPMENT PATTERNS
Dx(LF)of i
feed force
D y(HF) of * thrust force
D X(HF) ofi
feed force
DZ(HF) of -main cutting force
\ * t w
i
i
^
Figure 7.6 A typical relationship between tool wear and dispersion patterns
155
7.4 ARCHITECTURE OF NEURAL NETWORKS
7.4.1 Neural Network Techniques Used
By imitating the computational architecture of the human brain and implementing it
into software/hardware, neural networks are capable of learning to recognise non
linear and complicated input-output relationships. Back-propagation (BP) is the most
widely used learning algorithm for multilayered feed-forward neural networks [96-
98]. B P neural networks can be used to attack any problems that require pattern
mapping, i.e., given the input pattern, the network produces the associated output.
Once the network has learned the pattern mapping from the input-output training set,
for any new or previously unpresented input it will be capable of producing an output
pattern based on the knowledge derived from the recognised input-output relationship.
The typical BP neural network employs one hidden layer of artificial neurons fully
connected through weights to the input and output layers. The significance of using
the hidden layer is that it allows the non-linear mappings between input and output
patterns. Shown in Figure 7.7 is a three-layer neural network with N inputs, M
hidden neurons and L output neurons.
In Fig. 7.7, the forward propagation takes place first after an input pattern is presented
at the input layer. The errors, i.e. the differences between the output pattern (Oj, j=l,
2, ..., L) and target pattern (Tj, j=l, 2, ..., L) are calculated for all neurons in the
output layer and propagated back through the network to update their coming weights.
Next an error value is calculated for all the neurons in the hidden layer and the weights
are adjusted for all interconnections coming from the input layer. This process is
156
repeated until the output pattern is close enough to the target partem or until the error is
within the convergence criterion deterrnined in advance. The bias is connected to each
neuron in the hidden and output layers and is used to adjust the activation threshold of
the artificial neurons during the training process.
input pattern
Xi
X2
input layer
X N
hidden layer
output layer
output pattern
O 1
O,
o.
target pattern
Ti
Figure 7.7 Back-propagation neural network with one hidden layer
The function of a neuron in the hidden layer and output layer is shown in Figure 7.8.
157
Xi
Neuron j in hidden layer
Bias
Figure 7.8 The function of artificial neuron
In Fig. 7.8, the total input Xi,..., X n to Neuron j is a weighted linear summation,
i.e.,
Zj = IW j i . X i + Bj i=l
(7.1)
where Wji is the weight from the /th input to Neuron j and Bj is the bias of the Neuron
j. Then a real output value from Neuron j is activated by a non-linear transfer function
f(Zj),
Y j = f(Zj) = f ( £ w j i . X i + B j ) i=l
(7.2)
158
There are many transfer functions which might be implemented in the B P neural
network, which only requires the functions be differentiable everywhere [97]. Figure
7.9 describes three commonly used non-linear transfer functions.
NAME
S l-O
z
2 CC LU LL
CO
cr
Sigmoid
fl(Z)i
f/?\
Hz)- 1
n.o
z
0.0
1 + e"z
TanH (Hyperbolic Tangent]
ffZ) -Ai.o
r -
ez
1-1.0
- ez
T(ZJ-~z ~-z e + e
Sine
f(Z)Ai.o
X z
f(z) =
-1.0
SIN(z)
Figure 7.9 Description of three basic transfer functions
7.4.2 Selection of Input Features
The appropriate selection of input features is vital to the success of neural networks
and depends on the thorough understanding of the problem in question. A s described
before, chip breakability and surface finish change with tool wear progression, and
therefore the features selected should be sensitive to the specific wear types which
influence the chip breakability and surface finish. Major flank wear and crater wear
(crater depth) significantly change the tool configuration/geometry, thus resulting in
the variation of chip breakability, while minor flank wear is the dominant factor
influencing the surface quality of a finished product. Since the dispersion patterns
derived from multivariate time series models are sensitive to the rate of the above-
mentioned wear types, four associated dispersion features, viz., the low frequency
159
(LF) dispersion DX'(LF) of the feed force, and the high frequency (HF) dispersions
DX'(HF) of the feed force, Dy'(HF) of the thrust force and DZ'(HF) of the main
cutting force are selected as the input features representing effects of tool wear on chip
breakability and surface finish.
In addition, an initial condition for assessing chip forming patterns is required. As
described before, assessment of chip shapes/sizes for unworn tools is achieved in
terms of a fuzzy membership value, p. This value is selected as the fifth feature.
T w o machining parameters, cutting speed and feed, are also selected as the input
features due to their close relationship with chip breakability and surface finish.
Figure 7.10 shows the schematic diagram of feature selection, and Table 7.3 lists the
input-output data under Training Cutting Conditions 1 to 3 as representatives. The
input-output data under all the other cutting conditions are given in Appendix D.
On-line Measurement of 3-D Dynamic Cutting Force
1 Multivariate Time Series Modelling
l 1 Dispersion
Patterns
4 features
Cutting Speed
and Feed Rate
_ ! ' *
2 features
f
Machining Conditions
4 Basic Chip Database
J Initial Prediction
of Chip Breakability
r
Neural Network
1 feature
Figure 7.10 Feature selection for neural network
160
Table 7.3 The training data under Training Cutting Conditions 1-3
•i <3
§
8 1 u
1 5
INPUT FEATURES
Feed Speed D1(LF) DL(HF) rx(HF) D£(HF) l i - i i M-i(k)
(mm/tev) (m/min) (min ) (min ) (min ) (min )
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.06
0.06
0.06
0.06
0.06
0.06
0.06
115
115
115
115
115
115
115
115
145
145
145
145
145
145
145
145
145
145
145
145
145
145
-4.91
-3.89
-3.03
-1.15
0.73
2.61
4.49
7.78
-5.29
-3.11
-1.01
1.01
3.11
5.13
7.48
-8.50
-5.80
-3.20
-0.70
1.90
4.40
7.30
3.22
2.52
1.92
0.62
-0.68
-1.98
-3.28
-5.62
3.83
2.31
0.78
-0.75
-2.27
-3.92
-5.32
4.06
2.71
1.41
0.16
-1.14
-2.39
-3.84
1.50
1.31
1.14
0.78
0.42
0.06
-0.30
-0.95
5.22
3.84
2.54
1.29
-0.01
-1.26
-2.71
4.89
3.88
2.90
1.97
0.99
0.05
-1.04
1.47
1.15
0.87
0.27
-0.33
-0.93
-1.53
-2.61
4.01
2.79
1.56
0.34
-0.89
-2.21
-3.34
2.87
2.06
1.29
0.55
-0.23
-0.97
-1.84
0.25
0.25
0.30
0.32
0.35
0.41
0.58
0.46
0.25
0.25
0.27
0.31
0.41
0.47
0.45
0.23
0.23
0.24
0.27
0.32
0.41
0.43
OUTPUTS
Ra i ^0 ( k ) (um)
0.25
0.30
0.32
0.35
0.41
0.58
0.46
0.42
0.25
0.27
0.31
0.41
0.47
0.45
0.42
0.23
0.24
0.27
0.32
0.41
0.43
0.41
1.09
1.14
1.23
1.32
1.78
1.99
2.35
2.68
1.04
1.12
1.27
1.64
1.88
2.08
2.27
1.00
1.06
1.14
1.35
1.73
1.92
2.13
In Table 7.3, pi(k) and Po(k), k = 1, 2,..., represent the input and output of chip
breakability respectively, and
1 6 1
,Ui(0) = predicted value from the basic chip database, and
Pi(k) = p0(k-1) (7.3)
Equation 7.3 assumes that no sudden change in chip breakability will occur as tool
wear is normally a process of gradual progression. It is therefore reasonable to use
the output of chip breakability from the neural network at previous time interval Po(k-
1) as the value of current input Pi(k), with the initial chip breakability pi(0) from the
established basic chip database for unworn tools.
7.5 ANALYSIS OF RESULTS FROM NEURAL NETWORKS
7.5.1 Training Strategy with Neural Networks
The objective of using neural networks is to predict the development patterns of chip
breakability and surface finish at different wear states. Therefore, the input data
should be presented to the neural network by group in order to learn the development
trend of chip breakability and surface finish during the process of tool wear. The
training task in this work is to have the neural network learn the mappings from the
given input patterns to the desired output patterns under the Training Cutting
Conditions 1 to 6 (Table 7.1) which contain 41 samples in total. A s the system under
investigation contains only 7 inputs and 2 outputs, one hidden layer is sufficient to
establish an effective neural network [98].
How to select the optimum number of hidden neurons is a critical yet complicated
issue in Back-propagation networks [99-100], In this work an experimental approach
was taken to the selection of the number of hidden neurons. The training procedure
162
started with 7 and ended with 14 hidden neurons in line with the number of
inputs/outputs. Due to lack of knowledge about what transfer function would give the
best performance for the problem in question, three common transfer functions, i.e.
sigmoid, hyperbolic tangent (TanH) and sine, are used in each ttaining process. As
the training time is not important in the off-line stage for the size of neural networks
concerned, the training process was stopped only when no further improvement was
observed.
7.5.2 Analysis of Results for Training and Testing Effects
A Macintosh-based package, NeuralWorks Professional II, was used to establish the
neural network. Figure 7.11 shows the R M S errors for all the neural networks
trained with different number of hidden neurons and different transfer functions. It is
seen that with 12 hidden neurons is the minimum R M S error achieved for all three
transfer functions. This is in agreement with [12] where the experimental results
show that up to a certain number, further increase of hidden neurons does not always
lead to a better performance of neural networks. From Fig. 7.11, it can also be
concluded that among three transfer functions, TanH gives the minimum R M S errors,
i.e. 0.014 which should be thought as sufficiently close to zero [98].
The training effects of the selected 7-12-2 (i.e. 7 inputs, 12 hidden neurons & 2
outputs) neural network with the TanH transfer function are shown in Figures 7.12 &
7.13 for chip breakability and surface finish, respectively. The results indicate that the
selected neural network is capable of learning the mappings between the given input
patterns and the desired output patterns with quite acceptable accuracy.
163
0.04
0.03-
o fc. Ul
OT 0.02-
K
0.01 -
0.00
Sigmoid function
TanH function
Sine function
— « 1 • r — « 1 — - 1 1 • 1 • 1 • r—•« r -
6 7 8 9 10 11 12 13 14 Number of Hidden Neurons
15
Figure 7.11 Effect of the number of hidden neurons for three transfer functions
Time (min)
(a) Training Cutting Condition 1
Time (min)
(b) Training Cutting Condition 2
164
1.0
0.8 -
>. = 0.6 -.o
0.0
V • 145 m/min f = 0.06 mm/rev d = 0.5 m m
-1 5 Time
o targeted output x trained output
V = 205 m/min f = 0.06 mm/rev d = 0.5 m m
•
\ ^
1 1 r — i r-l
o targeted output x trained output
L T — • — i — ' — r ™ ~ ' 10 15
(min)
(c) Training Cutting Condition 3
4
Time 6 8 10
(min)
(d) Training Cutting Condition 4
1.0
0.8 -
1 0.6 K «
m 0.4
Q.
0.2
0.0
V = 170 m/min f = 0.1 mm/rev d = 0.5 m m
•
• 1 '
o targeted output
x trained output 1 , , ,—1
V = 160 m/min f = 0.15 mm/rev d = 0.5 m m
— i — | - — i — i — i — 1
o targeted output x trained output H — i — — i — i — r — 1
5
Time 10 15 0
(min)
(e) Training Cutting Condition 5
4
Time 8 10 6
(min)
(f) Training Cutting Condition 6
Figure 7.12 Evaluation of training effect of chip breakability
165
o.v
?2.5
• tt 2.0 a m •
3 o DC . c 1.0 o •
W 0.5
0.0
V f = d =
•
= 115 = 0.1 = 0.5
— r
m/min _ mm/rev ^^^m nm j ^ ^
o targeted output x trained output
!—1 1 1 1 10 Time
20 (min)
30
V = 145 m/min f = 0.1 mm/rev d = 0.5 m m
o targeted output x trained output
5 Time
— r ~ 10
(min)
15
(a) Training Cutting Condition 1 (b) Training Cutting Condition 2
3.0
f 2 . 5
2.0 -
m
c £ O) 3 O CC • o a 3 (0
1.5 -
1.0
0.0
V = 145 m/min f = 0.06 mm/rev d = 0.5 m m
• 1 r"
o targeted output x trained output
1 1 — — « i J
V = 205 m/min f = 0.06 mm/rev d = 0.5 m m
5 10 15 0
Time (min)
(c) Training Cutting Condition 3
-r 2
o targeted output x trained output
4 Time
-r 8 10 6
(min)
(d) Training Cutting Condition 4
166
o.u
?2.5
• DC 2.0 • • • ^ 1.5 a 3 o DC
1.0, • o • t to 0-5
0.0 C
V = 170 m/min f = 0.10 mm/rev d = 0.5 m m
^^6
^^
-
o targeted output x trained output
• 1 • 1 '• 9
) 5 10 15
V = 160 m/min f = 0.15 mm/rev d = 0.5 m m
•
^ ^ J J
^0^
^r _^*^t^^^
'
-
o targeted output x trained output
• 1 • 1 • i i [
0 2 4 6 8 10
Time (min)
(e) Training Cutting Condition 5
Time (min)
(f) Training Cutting Condition 6
Figure 7.13 Evaluation of training effect of surface finish
Four groups of Testing Cutting Conditions (Table 7.1), which were not used in
training of the neural network, are used to test the performance of this 7-12-2
network. Figures 7.14 & 7.15 show the comparisons between the actual outputs and
the predicted network outputs for chip breakability and surface finish, respectively.
Although the results are not as good as those during the training shown in Figs. 12 &
7.13, they should be considered accurate enough to describe the in-process
relationship between chip breakabiUty/surface finish and tool wear states under finish-
machining conditions. Therefore, the results from the established neural network,
together with comprehensive tool wear estimation, would give a quite satisfactory on
line assessment of machining performance in finish-machining.
167
1.0
0.8
>« t-p
2 0.6
a
o
0.2
0.0
,
-
r-~-
V = 140 m/min f m 0.12 mm/rev d = 0.5 m m
* * *
- i — - i — - i — '
? ***
o actual output
x network output 1 1 » 1 — '
V = 165 m/min f = 0.06 mm/rev d = 0.5 m m
' 1 » — '
o actual output
x network output 1 1 1 1—'
10 (min)
15 Time
(a) Testing Cutting Condition 1
5 10 15
Time (min)
(b) Testing Cutting Condition 2
1.0
0.8 -
>< ** S 0.6 m m
O 0.2 h
0.0
V = 190 m/min f = 0.06 mm/rev d = 0.5 m m
-
r T"
o actual output
x network output [_ ^j
5 Time
10 (min)
(c) Testing Cutting Condition 3
V = 130 m/min f = 0.15 mm/rev d = 0.5 m m
o actual output
x network output — I —
5 10 15
Time (min)
(d) Testing Cutting Condition 4
Figure 7.14 Neural network performance in predicting patterns of chip breakability
168
3.0
| 2.5
2.0
1.5 3 O DC
V = 140 m/min f - 0.12 mm/rev d =s 0.5 m m
8 i-o! • 3 10 0.5
0.0 1 ,
o actual output
x network output L 1 • r—'
V = 165 m/min f = 0.06 mm/rev d = 0.5 m m ^
^^^^
1 1—
o actual output
x network output \ 1- 1 |
Time 10
(min) 15 0
(a) Testing Cutting Condition 1
5 Time
10 (min)
15
(b) Testing Cutting Condition 2
3.0
I " • K 2.0 m m 0 c
Rough
•urface b
"' 0.5
0.0 (
V = 190 m/min i- f = 0.06 mm/rev
d = 0.5 m m
^^r
>
o actual output
x network output
) 5 10
Time (min)
(c) Testing C utting Condition 3
V = 130 m/min f = 0.15 mm/rev d = 0.5 m m
jg^
^C^^^
•
o actual output
x network output
0 5 10 15
Time (min)
(d) Testing Ci atting Condition 4
Figure 7.15 Neural network performance in predicting patterns of surface finish
169
7.6 EXTENSION : FUTURE DIRECTIONS AND DEVELOPMENTS
7.6.1 Real-time Application of Chip Control and Tool Wear
Estimation in Automated Machining Systems
Successful applications of chip control and tool wear estimation to the actual
automated machining systems rely heavily on the support of a powerful and
comprehensive software tool for the purpose of real-time monitoring/control. It is
suggested that this software, incorporated with an appropriate hardware, should have
the multiple functions of handling large scale database and knowledge base, executing
complicated scientific calculation, processing logic reasoning and knowledge
synthesising, and perforating real-time monitoring and control.
As an example, Figure 7.16 shows a possible project for an integrated system of chip
control and tool wear estimation, based on the software "G2 Real-time Expert
System" * which may meet the above-mentioned requirements.
* "G2 Real-time Expert System" is developed by Gensym Corporation, Cambridge, M A
02140, U.S.A. This software has recently become available commercially.
170
Basic Machining Experiments for Setting up the Standard Reference for Chip Control Assessment
MJ Chip Breaking
Chip Shapes /Sizes
3-D Chip Flow
r~r~i
Extensive Machining Experiments on the
Effect of Each Machining Parameter
on Chip Breaking and Forming Behaviours
Chip Control Database
1. Expert's Knowledge and Experience
2. Machining Theories 3. Empirical Rules
Off-line Guidances
1. Chip Breaking Prediction
2. Machining Performance Assessment
3. Machining Parameter Selection
On-line Assessment
1. Chip Forming Patterns
2. Surface Finish Development
3. Comprehensive I Tool Wear I Estimation I
On-line Control
1. Tool Change Policy
2. Dimensional Compensation
3. Machine-tool Operation
l_
I
Machining Experiments on 1. Surface
Finish 2. Power
Consumption 3. Tool Life
Chip Control Knowledge Base
I Machining Performance Database
G2 REAL-TIME EXPERT SYSTEM
~l Predicting Chip Breaking and Forming
Behaviour and the Primary Assessment of Machining Performance under the
Conditions of Cutting with Unworn Tools
I Integrated Outputs of
Chip Control and Tool Wear Estimation
I Comprehensive Tool Wear Estimation
I
Neural Network for Predicting the Dynamic Chip Forming Patterns
and Surface Finish
Development with Tool Wear Progression
I Multiple Dispersion Analysis
I Multivariate Time Series Modelling
— t - - — 3-D Dynamic Cutting Force and 3-D Vibration Measurements
-J
Figure 7.16 Proposed schematic diagram of an integrated system for chip control and tool wear estimation based on the G 2 Real-time Expert System
1 7 1
7.6.2 Developing a Self-organising Neural Network
Back propagation neural network techniques have been successfully used to predict
dynamic chip forming patterns and surface finish with tool wear progression under the
selected machining conditions. However, the major drawback is that the back
propagation algorithm is based on a "supervised" learning strategy, which needs
desired output patterns in each case. Apparendy, it is not feasible to conduct extensive
experiments to find the target patterns for numerous combinations of work materials
and tool configurations/geometries. Therefore, a strategy of "learning by self-
organising" should be developed to replace currently used "learning by being shown".
Self-organising learning is much more challenging, however, it can be used to resolve
very complicated problems such as aircraft engine monitoring [101]. The significance
of a self-organising neural network is its ability to adapt successfully to the
environments where rules m a y change unpredictably, that is, the ability to adapt
through direct confrontation with its "experiences" without a teacher to "supervise"
[102]. Therefore, employing a self-organising neural network is highly suggested for
future work in order to develop an integrated machining process monitoring/control
system which is applicable to the actual workshop environments.
7.7 SUMMARY AND CONCLUDING REMARKS
(1) The interaction of chip forming patterns with tool wear progression is extremely
complicated. N o effective theories at present can describe their interrelationship
analytically. Thus neural networks, which rely largely on input-output data,
have proven effective for the problem under investigation in this chapter.
172
(2) Based on the previous work described in Chapters 2 and 5, where the chip
breaking and forming behaviour when machining with unworn cutting tools
and comprehensive tool wear estimation can be predicted, an approach for the
in-process assessment of chip breakability and surface finish is achieved by
using neural network techniques.
(3) With the appropriate selection of a neural network structure and non-linear
transfer function, the mapping between the given input and output patterns can
be quite precisely achieved through training. Thus predictions of chip
breakability and surface finish for any new input data, which were not used in
training and may come from different cutting conditions, can be achieved with
quite acceptable accuracy.
(4) Although the results of this work are only for a flat-faced tool, they can be
extended to more complicated tool configurations by using the methodology
presented in this chapter. The method may also be extended to rough-
machining conditions where power consumption may have to be included in the
neural networks because of its critical effect on machining performance with
tool wear progression.
(5) The integration of the prediction of chip forming pattern, comprehensive tool
wear estimation, as well as surface finish, through the use of neural networks
provides an effective means for on-line assessment of machining performance
in automated machining systems.
173
CHAPTER 8
CONCLUSIONS
Through theoretical analysis and experimental research, this thesis presents a feasible
means for achieving effective chip control and comprehensive tool wear estimation,
and for further integrating these two vital concerns into an overall on-line assessment
of machining performance. The thesis has resulted in a better understanding of chip
control and tool wear estimation in automated machining systems, including the
following aspects:
(a) quantitative chip breakability prediction when machining steels;
(b) machining performance assessment including chip control, surface finish
and power consumption;
(c) tool chip breaker design with the evaluation of three-dimensional chip flow;
(d) comprehensive tool wear estimation, viz. major flank wear, crater wear,
minor flank wear and groove wear at the minor cutting edge, in oblique
machining;
(e) dynamic chip forming behaviour assessment with tool wear progression.
The major findings of this work are summarised below.
174
8.1 A NEW APPROACH FOR THE QUANTITATIVE PREDICTION OF CHIP BREAKABILITY AND CHIP FORMING PATTERNS
Prediction of chip breakability/chip forming patterns with sufficient accuracy for
arbitrary machining conditions is necessary because no suitable theories for chip
breaking in machining are available for use at the shop floor level. Almost every
theory that has been presented to date seems to be either of "academic" nature or of a
descriptive type with no specific applicable methods recommended. This is attributed
to the extremely complex nature of chip formation patterns with varying tool
configuration/geometry features and their interactions with work materials under
various cutting conditions.
As a new approach, a fuzzy-set mathematical model, in conjunction with a knowledge
database and the corresponding set of knowledge rules, has been implemented into an
expert system for predicting the "probable" levels of chip breakability in terms of a
fuzzy membership value.
The significance of the method for predicting chip breakability presented in this thesis
lies in that chip breakability can be quantified through a fuzzy rating system according
to the chip shapes/sizes produced, and further predicted through a fuzzy-set model
with very reasonable accuracy under any given set of input machining conditions,
including work materials, chip breaker configurations, tool geometries and cutting
conditions. Therefore, this method is suitable for application in an industrial
environment in automated machining systems.
The method for predicting chip breakability has been further extended to incorporate
the chip forming patterns and chip acceptability (for chip disposal) to give an
175
integrated evaluation of chip control in machining. Based on this, a predictive expert
system for machining performance assessment with chip control as a major criterion
has been developed with due consideration of surface finish and power consumption.
Thus, the method presented may be used as a basis for the assessment of "total
machinabihty" in automated machining systems.
8.2 A SCIENTIFICALLY-BASED METHOD FOR
THE EFFECTIVE DESIGN OF CHIP BREAKERS
The apparent "try and see" methods adopted by cutting tool manufacturers in the
design of chip breakers have to date resulted in hundreds of different chip breaker
configurations. Most of them, however, seem to produce desired chips only under
specific cutting conditions for selected work materials. W h e n these conditions are
changed and/or different work materials are used, chip breakability often becomes
quite different
With the aim of developing a scientifically-based method for designing groove-type
chip breakers, a knowledge-based system has been developed based on the analysis of
three-dimensional chip flow in oblique machining for a wide range of machining
conditions covering various work materials, cutting conditions, tool restricted contact
lengths, chip-breaker groove styles/sizes and tool geometries.
The merit of this method lies in that the optimum design of a chip breaker
configuration can be achieved based on the criterion of efficient chip breaking at
reduced power consumption. All the major factors which influence the effectiveness
of chip breaker configurations are quantitatively evaluated by using an experimentally-
based database and an integrated knowledge-base. Therefore, the methodology
176
presented in this thesis may provide an effective alternative to the traditional method of
designing groove-type chip breakers and also can be used as a guideline for cutting
tool manufacturers when designing any other types of tool chip breakers.
8.3 A NEW STRATEGY FOR COMPREHENSIVE
TOOL WEAR ESTIMATION
The conventional method for tool wear estimation is set only based on two common
types of wear, i.e., major flank wear and crater wear. N o one has made the effort to
include the wear states on the minor flank face which are crucial to assure surface
quality and dimensional accuracy. Especially in finish-machining, estimation of major
flank and/or crater wear alone is no longer adequate to describe the overall wear states.
This work a first attempt to develop a new strategy for comprehensive tool wear
estimation, including major flank wear, crater wear, minor flank wear and groove
wear at the minor cutting edge. Multivariate time series modelling techniques are
successfully applied to decompose data sampled from the machining process, such as
the 3-D dynamic cutting force and 3-D tool vibration, in terms of dispersions and/or
multiple dispersions. The success of employing dispersion analysis to comprehensive
tool wear estimation is attributed to its ability to single out the signal ingredients
sensitive to particular types of wear to be estimated.
The strategy of comprehensive tool wear estimation presented in this thesis may
provide a feasible means for on-line tool wear monitoring in automated machining
systems to assure product quality. The thesis also emphasises, in contrast to currently
used criteria, that in finish-machining tool change policy or optimal cutting conditions
177
should be determined based on wear states on the minor flank face because it is likely
that they reach critical points earlier than those on the major flank face and rake face.
8.4 A NEW ATTEMPT AT ESTIMATING CHIP FORMING PATTERNS WITH TOOL WEAR PROGRESSION FOR ON-LINE ASSESSMENT OF MACHINING PERFORMANCE
Chip control and tool wear estimation are two major concerns in automated machining
systems. They are closely interrelated during the machining process. The former
facilitates the safety of the machining operation, the maintenance of good surface
finish on the machined work surface and the convenience of chip disposal, while the
latter determines the tool change policy and the quality control strategy, especially in
the finish-machining processes.
Since the present theories and knowledge about metal machining are all based on ideal
machining conditions with unworn cutting tools, they are inadequate to quantitatively
describe the actual chip forming patterns with tool wear progression. W h e n tool wear
formed during the machining process becomes considerable, the chip forming patterns
vary significantly, or even totally differ, from those predicted for unworn cutting
tools, due to the changes in machining process characteristics and tool
configurations/geometries. Work has been seldom reported on studying the dynamic
chip forming behaviour with tool wear progression, mainly due to the extreme
complexity involved.
As a new attempt, neural network techniques are introduced to model the dynamic
interrelationship between chip forming patterns and tool wear development.
Considering that in finish-machining surface quality is of great importance to the
178
product being machined, dynamic prediction of surface roughness, which depends
primarily on wear states on the minor flank face for a worn tool, is also included.
The advantage of this new approach is that the previously developed methods, such as
quantitative prediction of chip breakability and comprehensive estimation of tool wear,
can be effectively integrated based on the same signal processing system, which
would greatly facilitate meeting the real-time requirement Therefore, the established
tool wear monitoring system, once implemented using the results developed from the
neural network, provides a feasible means for on-line assessment of machining
performance, including chip breaking/forming patterns, surface finish and overall tool
wear progression, which may give a satisfactory evaluation of dynamic machining
performance in finish-machining processes.
8.5 SUGGESTION FOR FUTURE W O R K
Due to the heavy dependence of tool wear on work materials and tool
configurations/geometries, it is impossible to develop a single on-line tool wear
recognition algorithm which is effective for arbitrary machining conditions. In order
to extend the method developed in this thesis for tool wear estimation to suit a wide
range of machining conditions, further efforts are required. It is, however, not
feasible to conduct extensive tool wear experiments due to the fact that tool wear
experiments are costly and time-consuming. A n alternative m a y be possible by
utilising and synthesising the present knowledge and database about tool wear and
machining operation, with the aid of an expert system.
179
The need for developing the techniques of artificial intelligence or expert systems for
on-line tool wear monitoring has been highlighted in recently published survey papers
about tool wear estimation [5,103-104]. Such an idea was presented in a recent work
[105] by introducing an expert system-supported tool wear monitoring strategy based
on a knowledge base. The basic thrust is based on the assumption that the distribution
of the forces acting on different tool faces (see Figure 5.9) depends mainly on tool
cutting edge angle, tool rake angle, chip breaker type as well as the three-dimensional
chip flow. The established multiple dispersion patterns derived from A R M A V models
of 3-D dynamic cutting force can be selected to establish the standard tool wear
recognition algorithm. W h e n machining conditions are different from the standard
ones, a set of tool wear decision-making rules may be determined to modify tool wear
recognition algorithm according to the variations of force distribution acting on
different tool faces.
However, greater efforts are needed in analysing the complicated relationship between
different force distributions on tool faces and multiple dispersion patterns, and in
synthesising the knowledge that could be used to develop tool wear decision-making
rules. Figure 8.1 proposes a possible scheme for developing such a tool wear
monitoring system.
180
Knowledge Base for Predicting 3-D Chip Flow
I
Tool geometry Chip Breaker Type
I Distribution Models of the Forces acting on Different Tool Faces
I Expert System for Determining Tool Wear
Decision-Making Rules
Standard Tool Wear Recognition Algorithm
On-line Measurements of 3-D Dynamic Cutting Force under Arbitrary Machining Conditions
i Determining Tool Wear Recognition Algorithm
Off-line •1
Dispersion Analysis Based on A R M A V Model
On-line Tool Wear Monitoring
Figure 8.1 A proposed schematic diagram for an expert system-supported tool wear estimation system
181
REFERENCES
1. J. F. Kahles, "CIRP Technical Reports : Machinability Data Requirements for
Advanced Machining Systems", Ann. CIRP, 36(2), pp. 523-529, (1987).
2. I. S. Jawahir, "A Survey and Future Predictions for the Use of Chip Breaking
in Unmanned Systems", Int. J. Adv. Mfg. Tech., 3(4), pp. 87-104, (1988).
3. D. Li and J. Mathew, "Tool Wear and Failure Monitoring Techniques for
Turning - A Review", Int. J. Mach. Tools Manufact., 30, No. 4, pp. 579-598
(1990).
4. J. Tlusty, "A Critical Review of Sensors for Unmanned Machining", Ann.
CIRP, 32(2), pp. 563-572 (1983).
5. H.K., Toenshoff, et. al., "Developments and Trends in Monitoring and
Control of Machining Processes", Ann. CIRP, 37(2), pp. 611-622 (1988).
6. M., Shiraishi, "Scope of In-process Measurement, Monitoring and Control
Techniques in Machining Processes-Part 1 : In-process Techniques for Tools",
Precision Engineering, 10(4), pp. 179-189 (1988).
7. S.Y., Liang and D.A., Dornfeld, "Detection of Cutting Tool Wear Using
Adaptive Time Series Modeling of Acoustic Emission Signal", ASME/WAM,
Sensors for Manufacturing, PED - Vol. 26, pp. 27-38 (1987).
8. S.Y., Liang and D.A., Dornfeld, "Tool Wear Detection Using Time Series
Analysis of Acoustic Emission", /. of Eng. for Ind., Ill, pp. 199-205
(1989).
9. S.M, Pandit, H. Susuki and C H . Kahng, "Application of Data Dependent
Systems to Diagnostic Vibration Analysis", /. Mechanical Design, 102, 233-
241 (1980).
10. S.M. Pandit, and S. Kashou, "A Data Dependent Systems Strategy of On-line
Tool Wear Sensing", /. of Eng. for Ind., 104, pp. 217-223 (1982).
182
11. S. Rangwala and D. Dornfeld, "Sensor Integration Using Neural Networks for
Intelligent Tool Condition Monitoring", ASME Trans., J. of Eng. for Ind.,
112, pp. 219-228 (1990).
12. D.A. Dornfeld, "Neural Network Sensor Fusion for Tool Condition
Monitoring", Ann. CIRP, 39(1), pp. 101-105 (1990).
13. Z.J. Yuan, et. al., "In-Process Detection of Tool Wear and Breakage by
Autocorrelation Coefficient of Dynamic Cutting Forces in Turning", Proc. 4th
Int. Conf. on Manuf. Eng., Brisbane, Australia, pp. 201-205 (1988).
14. A.M. Petrie, T.S. Sifra and A. Ismail, "The Development of An Automated
System for On-line Tool Wear Monitoring", Proc. of 1st Int. Machinery
Monitoring & Diagnostics Conference, U.S.A., pp. 546-552 (1989).
15. Y. Koren et. al., "Monitoring Tool Wear Through Force Measurement", Proc.
of 15th NAMRC, pp. 463-468 (1987).
16. S.B. Rao, "Tool Wear Monitoring Through the Dynamics of Stable Turning",
ASME Trans., J. of Eng. for Ind., 108, pp. 183-190 (1986).
17. J. Vogel, S. Pamalingam and K. Stelson, "On Tool Post Vibration Monitoring
for Tool Condition Identification", Proc. of 14th NAMRC, pp. 281-285
(1986).
18. G. Chryssolouris, M. Guillot and M. Domroese, "Tool Wear Estimation for
Intelligent Machining", ASMEIWAM, Intelligent Control, D S C - Vol. 5, pp.
35-43 (1987).
19. F. Giusti, M. Santochi and G. Tantussi, "On-Line Sensing of Flank and Crater
Wear of Cutting Tools", Ann. CIRP, 36(1), pp. 41-44 (1987).
20. R. Teti, "Tool Wear Monitoring Through Acoustic Emission", Ann. CIRP,
38(1), pp. 99-102 (1989).
21. K. Danai and A.G. Ulsoy, "A Dynamic State Model for On-line Tool Wear
Estimation in Turning", ASMEIWAM, Sensors of Control of Manufacturing,
PED-Vol. 18, pp. 137-148 (1985).
183
22. K. Okushima, K. Hitomi and S. Ito, "A Study of Super-High Speed
Machining", Ann. CIRP, 13, pp. 399-410 (1966).
23. A.J. Pekelharing and R.A. Schuermann, "Wear of Carbide Tools - its effect of
surface finish and dimensional accuracy", The Tool Engineer, pp. 51-57 (Oct.
1953).
24. AJ. Pekelharing, "Finish Turning", Ann. CIRP, 8, pp. 112-120 (1959).
25. B.H. Lambert, "Two Years of Finish-Turning Research at the Technological
University, Delft", Ann. CIRP, 10, pp. 246-255 (1961-62).
26. AJ. Pekelharing, "Some Special Aspects of Carbide Tool Wear", International
Production Engineering Research Conference, A S M E , New York (1963).
27. S. Takata, and T. Sata, "Model Referenced Monitoring and Diagnosis ~
Application to Manufacturing System", Computers in Industry, 7, pp. 31-43
(1986).
28. K. Eman and S.M. W u , "A Feasibility Study of On-Line Identification of
Chatter in Turning Operations", J. of Eng. for Ind., 102, pp. 315-321 (1980).
29. X.D. Fang, Y. Yao and G. Arndt, "Tool Wear Estimation by Multidimensional
Autoregressive Spectral Analysis", Proc. 1990 Pacific Conference on
Manufacturing, Sydney & Melbourne, Australia, Vol. 2, pp. 834-841 (1990).
30. Y. Yao, X.D. Fang and G. Arndt, "Comprehensive Tool Wear Estimation in
Finish-Machining via Multivariate Time-Series Analysis of 3-D Cutting
Forces", Ann. CIRP, 39(1), pp. 57-60 (1990).
31. S.M. Pandit, "Analysis of Vibration Records by Data Dependent Systems",
Shock and Vibration Bulletin, 47(4), pp. 161-174 (1977).
32. Y. Yao, X.D. Fang and G. Arndt, "On-line Estimation of Groove Wear in the
Minor Cutting Edge for Finish-Machining", accepted for publication in Ann.
CIRP, 40(1) (1991).
184
33. P. Bandyopadhyay and S.M. W u , "Signature Analysis of Drilling Dynamics
for On-Line Drill Life Monitoring", ASMEIWAM, Sensors and Controls for
Manufacturing, pp. 101-110(1985).
34. M.S. Lan and Y. Naerheim, "In-Process Detection of Tool Breakage in
Milling", ASME Trans.. J. of Eng. for Ind., 108, pp. 191-197 (1986).
35. S.M. Pandit and S.M. W u , Time Series and System Analysis with
Applications. John Wiley, U S A (1983).
36. R.H. Jones, "Multivariate Autoregression Estimation using Residuals",
Applied Time Series Analysis. Academic Press Inc., pp. 139-162, (1978).
37. G. Chryssolouris and M. Domroese, "Sensor Integration for Tool Wear
Estimation in Machining", Proc. WAMIASME, Symposium on Sensor and
Control forManufacturing, pp. 115-123, (1988).
38. S. Rangwala and D. Dornfeld, "Learning and Optimization of Machining
Operations Using Computing Abilities of Neural Networks", IEEE Trans, on
Systems, Man and Cybernetics, 19(2), pp. 299-314 (1989).
39. K. Nakayama, "Chip Curl in Metal Cutting Process", Bulletin of the Faculty
of Engineering, Yokohama National Univ. 11, pp. 1-13 (1962).
40. K. Nakayama, "Chip Control in Metal Cutting", Bull. Japan. Soc. ofPrec.
Eng. 18(2), pp. 97-103 (1984).
41. B. Worthington and M. H. Rahman, "Predicting Breaking with Groove Type
Breakers", Int. J. MTDR, 19, pp. 121-132 (1979).
42. E. K. Henriksen, 'Balanced Design Will Fit the Chip Breaker to the Job : Chip
Breaker Dimensions are Critical in Taming Chips', American Machinist,
98(4), pp. 118-124(1954).
43. E. K. Henriksen, "Findings and Directions in Chip Breakers Research", Proc.
23rd Annual Meeting ofASTE, Los Angles, California (14-18 March 1955).
44. L. V. Colwell, "Predicting the Angle of Chip Flow for Single-Point Cutting
Tools", Trans. ASME , 76(2), pp. 199-204 (1954).
185
45. G. V. Stabler, "The Chip Flow Law and its Consequences", Proc. 5th Int.
MTDR Conf., pp. 243-251 (September 1964).
46. N. H. Cook, P. Jhaveri and N. Nayak, "The Mechanism of Chip Curl and its
Importance in Metal Cutting", Trans. ASME, pp. 374-380 (1963).
47. K. Nakayama, "A Study on Chip Breaker", Bulletin of JSME, 5(15), pp.
142-150 (1962).
48. K. Nakayama, "Pure Bending Test of Chip - An Approach to the Prediction of
Cutting Force", Bulletin of the Faculty of Engineering, Yokohama National
Univ., 12, pp.1-14 (1963).
49. K. Nakayama, "Origins of Side Curl of Chip in Metal Cutting", Bull. Japan
Soc. ofPrec. Eng., 6(3), pp. 99-101 (1972).
50. K. Nakayama and M. Ogawa, "Basic Rules on the Form of Chip in Metal
Cutting", Ann. CIRP, 27(1), pp. 17-21 (1978).
51. K. Nakayama and M.Arai, "The Breakability of Chip in Metal Cutting", Proc.
Int.Conf.onManuf.Eng., Melbourne, Australia, pp. 6-10 (1980).
52. K. Nakayama, M. Arai and T. Kondo, "Cutting Tools with Curved Rake Face
- A Means for Breaking Thin Chips", Ann. CIRP, 30(1), pp. 5-8 (1981).
53. B. Worthington and A. H. Redford, "Chip Curl and the Action of the Groove
Type Chip Former", Int. J. MTDR, 13, pp. 257-270 (1973).
54. B. Worthington, "The Effect of Rake Face Configuration on the Curvature of
the Chip in Metal Cutting", Int. J. MTDR, 15, pp. 223-239 (1975).
55. B. Worthington, "The Operation and Performance of a Groove-type Chip
Forming Device", Int. J. Prod. Res., 14(5), pp. 529-558 (1976).
56. W . Johnson, "Some Slip-line Fields for Swaging or Expanding, Indenting,
Extruding and Machining for Tools with Curved Dies", Int. J. Mech. Sci., 4 ,
pp. 323-347 (1962).
186
57. E. Usui and K. Hoshi, "Slip-line Fields in Metal Machining Which Involve
Centered Fans", Proc. Int. Prod. Eng. Res. Conf. ASME, Pittsburgh, pp.
61-71 (September 1963).
58. E. Usui, K. Kikuchi and K. Hoshi, "The Theory of Plasticity Applied to
Machining with Cut-Away Tools" Trans. ASME, pp. 95-104 (May 1964).
59. I. S. Jawahir, "The Chip Control Factor in Machinability Assessments : Recent
Trends", Journal of Mechanical Working Technology, 17, pp. 213-224
(1988).
60. I. S. Jawahir and P. L. B. Oxley, "Efficient Chip Breaking at Reduced Power
Consumption - An Experimental Analysis", Proc. 4th Int. Conf. on
Manufacturing Engineering, Brisbane, Australia, pp. 97-102 (May 1988).
61. I. S. Jawahir, "The Tool Restricted Contact Effect as a Major Influencing
Factor in Chip Breaking : An Experimental Analysis", Ann. CIRP, 37(1),
pp. 121-126 (1988).
62. I. S. Jawahir and P. L. B. Oxley, "New Developments in Chip Control
Research; Moving Towards Chip Breakability Predictions for Unmanned
Manufacture", Proc. Int. Conf. ASME - MI'88, ASME Publ., Atlanta, USA,
Vol. 1, pp. 311-320 (April 1988).
63 I. S. Jawahir, "On the Controllability of Chip Breaking Cycles and Modes of
Chip Breaking in Metal Machining", Ann. CIRP, 39(1), pp. 47-51 (1990).
64. C. L. Hsu, "An Expert System Approach for Selection and Optimization of
Machining Process Parameters", PhD Thesis, University of Illinois at Chicago,
U.S.A. (1985).
65. H. P. Wang and R. A. Wysk, "An Expert System for Machining Data
Selection", Comput. & Indus. Engg, 10(2), pp. 99-107 (1986).
66. F. Giusti et al, "COATS : An Expert Module for Optimal Tool Selection", Ann.
CIRP, 35(1), pp. 337-340 (1986).
187
67. I. S. Jawahir and X. D. Fang, "A Knowledge-based Approach for Improved
Performance with Grooved Chip Breakers in Metal Machining", Proc. 3rd Int.
Conf. on Advances in Manufacturing Technology, Singapore, pp. 130-144
(August 1989).
68 C. Spaans, "The Fundamentals of Three-Dimensional Chip Curl, Chip
Breaking and Chip Control", Doctoral Thesis, T H Delft, (1971).
69. W . K. Luk, "The Direction of Chip Flow in Oblique Cutting", Int. J. Prod.
Res., 10(1), pp. 67-76 (1972).
70. C. Y. Jiang, Y. Z. Zhang, Z. J. Chi, "Experimental Research of the Chip Flow
Direction and its Application to the Chip Control", Ann. CIRP, 33(1), pp. 81-
84 (1984).
71. H. T. Young, P. Mathew and P. L. B. Oxley, "Allowing for Nose Radius
Effects in Predicting the Chip Flow Direction and Cutting Forces in Bar
Turning", Proc. Inst. Mech. Engrs., 201(C3), pp. 213-226 (1987).
72. J. Wang and P. Mathew, "Predicting the Chip Flow Direction for Nose Radius
Tools under Oblique Machining Conditions", University of New South Wales
Report, Australia, 1988/IE/2 (1988).
73. V.LSisoev, Fundamentals of Metal Cutting and Cutting Tools. Mashgiz,
Moscow (1962).
74. J. S. Bator, "Power Reduction through Efficient Chip Control", Cutting Tool
Engineering, pp. 4-8 (July/August 1975).
75. A. Ber, S. Kaldor and E. Lenz, "New Concept in Chip-breaker Design Leads
to a Wider Range of Chip-Breaking", S M E Technical Paper, M R 79-307
(1979).
76. S. Kaldor, A. Ber and E. Lenz, "On the mechanism of chip breaking", Trans.
ASME, 101 pp. 241-249 (1979).
77. L. De Chiffre, "Cutting Tools with Restricted Contact", Int. J. Mach. Tool
Des. Res., 22(4), pp. 321-332 (1982).
188
78. L. De Chiffre, "Metal Cutting - Mechanics and Applications", D. Sc. Thesis,
Technical University of Denmark (1990).
79. I. S. Jawahir, "An investigation of Three-dimensional Chip Flow in Machining
of Steels with Grooved Chip Forming Tool Insert", Trans. NAMRC, Vol. 19,
pp. 222-231 (1991).
80. S. Rangwala and D. Dornfeld, "Integration of Sensors via Neural Networks
for Detection of Tool Wear States", ASMEIWAM, Intelligent and Integrated
Manufacturing Analysis & Synthesis, P E D - Vol. 25, pp. 109-120 (1987).
81. A. Del Taglia, et. al., "An Approach to On-line Measurement of Tool Wear by
Spectrum Analysis", Proc. of Int. MTDRC, pp. 141-148 (1976).
82. C Y . Jiang, Y.Z. Zhang and H.J. Xu, "In-Process Monitoring of Tool Wear
Stage by the Frequency Band-Energy Method", Ann. CIRP, 36(1), pp. 45-48
(1987).
83. T. Sata, et. al, "Learning and Recognition of the Cutting States by the Spectrum
Analysis", Ann. CIRP, 22(1), pp 41-42 (1973).
84. K.F. Eman and S.M. W u , "Present and Future Trends in Stochastic Analysis
of Cutting and Structural Dynamics", Proc. of 15th NAMRC, pp. 426-432
(1987).
85. W.R. DeVries, D.A. Dornfeld and S.M. W u , "Bivariate Time Series Analysis
of the Effective Force Variation and Friction Coefficient Distribution in Wood
Grinding", ASME Trans., J. of Eng. for Ind., 100, pp 181-185 (May 1978).
86. B. Mills and A. H. Redford, Machinability of Engineering Materials. Applied
Science Publishers (1983).
87. AJ. Pekelharing and C.A., Van Luttervelt, CIRP Terminology and Procedures
for Turning Research. Technical Secretary CIRP Group C, Delft, (1969).
88. B. W u , et. al., "The Dynamic Component of Cutting Force and its Application
in Cutting State Identification", /. of Nanjing Institute of Technology, 17,
No. 6 (I), pp. 55-65 (1987).
189
89. H. Akaike, "A New Look at the Statistical Model Indentification", IEEE Trans.
Automat. Contr., AC-19, pp. 716-723 (1974).
90. V. Solaja, "Wear of Carbide Tools and Surface Finish Generated in Finish
Turning of Steel", Wear, 2, pp. 40-58 (1958/1959).
91. M . C Shaw, and J.A. Crowell, "Finish Machining", Ann. CIRP, 13, pp. 5-22
(1965).
92. AJ. Pekelharing and HJ. Hovinga, "Wear at the End Cutting Edge of Carbide
Tools in Finish and Rough Turning", Proc. of 8th Int. MTDR Conf, pp. 643-
651 (1967).
93. S.M. Pandit, "Data Dependent Systems : Modeling Analysis and Optimal
Control via Time Series", PhD Thesis, University of Wisconsin, Madison, WI,
U.S.A. (1973).
94. X. D. Fang and I. S. Jawahir, "An Expert System Based on A Fuzzy
Mathematical Model for Chip Breakability Assessments in Automated
Machining", Proc. Int. Conf, ASME, MI'90, Vol. IV, Atlanta, Georgia, U.
S. A., pp. 31-37 (March 1990).
95. X. D. Fang and I. S. Jawahir, "On Predicting Chip Breakability in Machining
of Steels with Carbide Tool Inserts Having Complex Chip Groove
Geometries", to appear in /. Materials Processing Technology, 26, Issue 2,
selected from the papers in Proc. 7th International Conference on Computer-
Aided Production Engineering, Tennessee, USA, pp. 37-48 (1991).
96. D. Rumelhart and J. McClelland, Parallel Distributed Processing. Vol. 1, MIT
Press, Canbridge, M A (1986).
97. P.D. Wasserman, Neural Computing : Theory and Practice. Van Nostrand
Reinhold, New York (1989).
98. J.E. Dayhoff, Neural Network Architectures. Van Nostrand Reinhold, New
York (1990).
190
99. S.C Huang and Y.F. Huang, "Bounds on the Number of Hidden Neurons in
Multilayer Perceptrons", IEEE Trans, on Neural Networks, 2(1), pp. 47-55
(1991).
100. S.Y. Kung and J.N. Hwang, "Bounds on the Number of Hidden Neurons in
Multilayer Perceptrons", Proc. IEEE Int. Conf. on Neural Networks, Vol. I,
San Diego, California, pp. 363-370 (1988).
101. P.D. Hewitt, PJ.C Skitt and R.C. Witcomb, "A Self-organising Feedforward
Network Applied to Acoustic Data", New Development in Neural Computing,
ed. J.G. Taylor and CL.T. Mannion, Adam Hilger, Bristol and New York,
pp. 71-78(1989).
102. G.A. Carpenter and S. Grossberg, "The A R T of Adaptive Pattern Reconnition
by a Self-organising Neural Network", Artificial Neural Networks Concept
Learning, ed. J. Diederich, IEEE Computer Society Press, pp. 69-80 (1990).
103. Usui, E., "Progress of 'Predictive' Theories in Metal Cutting", JSME Int. J.,
Series HI, 31, No. 2, pp. 363-369 (1988).
104. Monostori, L., "Signal Processing and Decision Making in Machine Tool
Monitoring Systems", Proc. 16th NAMRC, pp. 277-284 (1988).
105. X. D. Fang and Y. Yao, "Expert System-Supported On-line Tool Wear
Monitoring", Proc. of International Conference on Artificial Intelligence in
Engineering - AIE'90, Sponsored by I.E.E.(UK), Kuala Lumpur, Malaysia,
pp. 77-84 (August 1990).
A-1
APPENDIX A STANDARD FUZZY MEMBERSHIP VALUES FOR CHIP BREAKABILITY
Table A-1 Standard fuzzy membership values p(x) for chip breakability Work Material: CS 1020 Chip Breaker: ENZ-type Tool Geometry : 0°, 5°, -6°, 90°, 60°, 0.8
Cutting Speed (m/min)
50
75
100
150
200
Depth of Cut (mm)
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
Feed (mm/rev)
0.06
0.49
0.51
0.53
0.58
0.43
0.45
0.23
0.25
0.27
0.41
0.40
0.44
0.16
0.17
0.19
0.23
0.26
0.31
0.11
0.13
0.16
0.16
0.21
0.26
0.09
0.11
0.13
0.16
0.18
0.22
0.1
0.40
0.41
0.47
0.51
0.41
0.43
0.21
0.23
0.23
0.26
0.38
0.40
0.19
0.19
0.19
0.21
0.28
0.35
0.16
0.16
0.16
0.31
0.24
0.30
0.13
0.13
0.13
0.13
0.20
0.26
0.2
0.56
0.58
0.46
0.38
0.40
0.43
0.49
0.50
0.45
0.38
0.38
0.40
0.48
0.48
0.45
0.35
0.37
0.39
0.47
0.47
0.41
0.41
0.36
0.38
0.43
0.43
0.37
0.27
0.32
0.33
0.3
0.77
0.77
0.73
0.43
0.41
0.40
0.74
0.73
0.73
0.43
0.40
0.40
0.71
0.71
0.71
0.42
0.40
0.38
0.68
0.68
0.69
0.41
0.39
0.37
0.64
0.64
0.65
0.37
0.35
0.33
0.4
0.82
0.86
0.83
0.46
0.42
0.41
0.83
0.80
0.80
0.45
0.41
0.40
0.75
0.80
0.77
0.45
0.41
0.39
0.72
0.77
0.74
0.43
0.40
0.38
0.68
0.73
0.70
0.40
0.36
0.34
0.5
0.86
0.91
0.87
0.47
0.44
0.42
0.87
0.83
0.83
0.46
0.43
0.41
0.82
0.84
0.81
0.45
0.43
0.40
0.80
0.81
0.80
0.43
0.42
0.40
0.76
0.77
0.75
0.40
0.38
0.36
A-2
Table A-2 Standard fuzzy membership values p(x) for chip breakability Work Material :K1040 Chip Breaker: ENZ-type Tool Geometry : 0°, 5°, -6°, 90°, 60°, 0.8
Cutting Speed (mAnin)
50
75
100
150
200
Depth of Cut (mm)
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
0.06
0.47
0.50
0.53
0.58
0.59
0.50
0.22
0.24
0.26
0.40
0.50
0.50
0.15
0.16
0.18
0.22
0.25
0.30
0.10
0.12
0.15
0.18
0.20
0.25
0.08
0.10
0.12
0.15
0.17
0.21
0.1
0.37
0.39
0.44
0.50
0.56
0.55
0.20
0.22
0.22
0.25
0.38
0.46
0.18
0.18
0.18
0.20
0.28
0.34
0.15
0.15
0.15
0.15
0.23
0.29
0.12
0.12
0.12
0.12
0.19
0.25
Keed (mm/rev)
0.2
0.68
0.66
0.49
0.37
0.39
0.50
0.64
0.62
0.49
0.36
0.38
0.40
0.60
0.59
0.49
0.34
0.36
0.38
0.42
0.43
0.37
0.30
0.35
0.36
0.38
0.39
0.32
0.26
0.31
0.32
0.3
0.76
0.76
0.73
0.43
0.40
0.40
0.73
0.73
0.73
0.43
0.40
0.40
0.70
0.70
0.70
0.42
0.39
0.38
0.67
0.67
0.68
0.40
0.38
0.37
0.63
0.63
0.64
0.36
0.34
0.33
0.4
0.81
0.85
0.82
0.45
0.41
0.41
0.71
0.82
0.79
0.45
0.41
0.40
0.74
0.79
0.76
0.44
0.40
0.39
0.71
0.76
0.73
0.43
0.39
0.38
0.67
0.72
0.69
0.39
0.35
0.34
0.5
0.85
0.90
0.86
0.45
0.43
0.42
0.83
0.86
0.82
0.46
0.43
0.41
0.81
0.83
0.80
0.44
0.42
0.40
0.79
0.80
0.79
0.43
0.41
0.40
0.75
0.76
0.74
0.39
0.37
0.36
A-3
Table A-3 Standard fuzzy membership values p(x) for chip breakability Work Material :EN25 Chip Breaker: ENZ-type Tool Geometry : 0°, 5°, -6°, 90°, 60°, 0.8
Cutting Speed (nymin)
50
75
100
150
200
Depth of Cut (mm)
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
Feed (mm/rev)
0.06
0.26
0.30
0.35
0.40
0.45 i
0.48
0.19
0.21
0.23
0.37
0.38
0.42
0.12
0.13
0.15
0.19
0.22
0.27
0.07
0.09
0.12
0.15
0.17
0.22
0.06
0.07
0.09
0.12
0.14
0.20
0.1
0.27
0.31
0.36
0.38
0.39
0.43
0.17
0.19
0.19
0.22
0.35
0.38
0.15
0.15
0.15
0.17
0.25
0.32
0.12
0.12
0.12
0.12
0.20
0.28
0.10
0.09
0.09
0.09
0.16
0.24
0.2
0.45
0.45
0.45
0.34
0.36
0.42
0.45
0.45
0.45
0.33
0.35
0.37
0.42
0.42
0.42
0.31
0.33
0.36
0.40
0.41
0.35
0.27
0.32
0.35
0.36
0.37
0.31
0.23
0.28
0.31
0.3
0.73
0.73
0.67
0.40
0.37
0.39
0.70
0.70
0.70
0.40
0.37
0.39
0.67
0.67
0.67
0.39
0.36
0.37
0.64
0.64
0.65
0.37
0.35
0.36
0.60
0.60
0.61
0.33
0.31
0.32
0.4
0.78
0.82
0.79
0.42
0.38
0.40
0.74
0.79
0.76
0.42
0.38
0.39
0.76
0.76
0.73
0.41
0.37
0.38
0.68
0.73
0.70
0.40
0.36
0.37
0.64
0.69
0.66
0.36
0.32
0.33
0.5
0.82
0.87
0.83
0.43
0.40 |
0.41
0.80
0.83
0.79
0.43
0.40
0.40
0.80
0.80
0.77
0.41
0.39
0.39
0.76
0.77
0.76
0.40
0.38
0.39
0.72
0.73
0.71
0.36
0.34
0.35
A - 4
Table A-4 Standard fuzzy membership values p(x) for chip breakability Work Material: CS 1020 Chip Breaker: ENA-type Tool Geometry : 0°, 5°, -6°, 90°, 60°, 0.8
Cutting Speed (m/min)
50
75
100
150
200
Depth of Cut ' (mm)
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
Feed (mm/rev) j
0.06
0.48
0.49
0.49
0.48
0.60
0.60
0.25
0.25
0.28
0.28
0.46
0.55
0.20
0.20
0.25
0.26
0.34
0.32
0.12
0.15
0.18
0.19
0.30
0.32
0.20
0.20
0.25
0.26
0.34
0.32
0.1
0.56
0.56
0.55
0.46
0.46
0.48
0.32
0.33
0.33
0.32
0.36
0.45
0.30
0.30
0.30
0.29
0.36
0.36
0.25
0.26
0.26
0.25
0.34
0.36
0.30
0.30
0.30
0.29
0.36
0.36
0.2
0.70
0.74
0.72
0.65
0.37
0.38
0.68
0.71
0.66
0.60
0.36
0.36
0.65
0.68
0.62
0.55
0.36
0.36
0.60
0.65
0.59
0.50
0.34
0.35
0.65
0.68
0.62
0.55
0.36
0.36
0.3
0.78
0.83
0.80
0.72
0.38
0.38
0.75
0.81
0.75
0.67
0.36
0.36
0.72
0.80
0.72
0.64
0.36
0.36
0.70
0.75
0.70
0.60
0.35
0.35
0.72
0.80
0.72
0.64
0.36
0.36
0.4
0.85
0.90
0.89
0.80
0.40 |
0.39 |
0.83
0.89
0.85
0.78
0.40
0.39
0.80
0.88
0.81
0.74
0.40
0.38
0.79
0.87
0.78
0.71
0.38
0.37
0.80
0.88
0.81
0.74
0.40
1 0.38
0.5
0.90
0.94
0.91
0.84
0.42
0.40
0.89
0.93
0.87
0.80
0.42
0.40
0.88 1 0.92
0.84
0.78
0.42
0.39
0.87
0.91
0.82
0.76
0.39
0.37
0.88
0.92
0.84
0.78
0.42
0.39
A-5
Table A-5 Standard fuzzy membership values p(x) for chip breakability Work Material :K1040 Chip Breaker: ENA-type Tool Geometry : 0°, 5°, -6°, 90°, 60°, 0.8
Cutting Speed (m/min)
50
75
100
150
200
Depth of Cut (mm)
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
0.06
0.48
0.48
0.48
0.48
0.61
0.53
0.25
0.25
0.28
0.28
0.54
0.52
0.20
0.20
0.25
0.25
0.35
0.43
0.11
0.14
0.17
0.18
0.30
0.32
0.20
0.20
0.25
0.25
0.35
0.43
0.1
0.47
0.46
0.46
0.46
0.55
0.58
0.32
0.33
0.33
0.32
0.35
0.52
0.30
0.30
0.30
0.29
0.35
0.36
0.25
0.26
0.26
0.25
0.33
0.36
0.30
0.30
0.30
0.29
0.35
0.36
Feed (mm/rev)
0.2
0.69
0.73
0.71
0.64
0.38
0.38
0.67
0.70
0.65
0.59
0.37
0.36
0.64
0.67
0.61
0.54
0.37
0.36
0.59
0.64
0.58
0.49
0.34
0.35
0.64
0.67
0.61
0.54
0.37
0.36
0.3
0.77
0.82
0.79
0.71
0.39
0.38
0.74
0.80
0.74
0.66
0.37
0.36
0.71
0.78
0.71
0.63
0.37
0.36
0.69
0.74
0.69
0.59
0.35
0.35
0.71
0.78
0.71
0.63
0.37
0.36
0.4
0.84
0.89
0.88
0.79
0.39
0.39
0.82
0.88
0.84
0.77
0.41
0.39
0.79
0.87
0.80
0.73
0.41
0.38
0.78
0.86
0.77
0.70
0.38
0.37
0.79
0.87
0.80
0.73
0.41
0.38
0.5
0.89
0.93
0.90
0.83
0.39
0.40
0.88
0.92
0.86
0.79
0.43
0.40
0.87
0.91
0.83
0.77
0.43
0.39
0.86
0.90
0.81
0.75
0.39
0.37
0.87
0.91
0.83
0.77
0.43
0.39
A - 6
Table A-6 Standard fuzzy membership values p(x) for chip breakability Work Material :EN25 Chip Breaker: ENA-type Tool Geometry : 0°, 5°, -6°, 90°, 60°, 0.8
Cutting Speed (m/min)
50
75
100
150
200
Depth of Cut (mm)
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
0.06
0.30
0.32
0.32
0.32
0.40
0.45
0.22
0.22
0.25
0.25
0.28
0.43
0.17
0.17
0.22
0.22
0.28
0.34
0.08
0.11
0.14
0.16
0.27
0.29
0.17
0.17
0.22
0.22
0.28
0.34
0.1
0.32
0.34
0.32
0.30
0.43
0.38
0.29
0.30
0.30
0.29
0.33
0.36
0.27
0.27
0.27
0.26
0.33
0.35
0.22
0.23
0.23
0.22
0.31
0.33
0.27
0.27
0.27
0.26
0.33
0.35
feed (mm/rev)
0.2
0.64
0.68
0.64
0.59
0.36
0.37
0.38
0.40
0.38
0.35
0.35
0.35
0.36
0.38
0.36
0.34
0.35
0.35
0.34
0.35
0.33
0.32
0.33
0.34
0.36
0.38
0.36
0.34
0.35
0.35
0.3
0.75
0.80
0.77
0.69
0.38
0.37
0.72
0.78
0.72
0.64
0.35
0.35
0.69
0.77
0.69
0.61
0.36
0.35
0.67
0.72
0.67
0.57
0.35
0.34
0.69
0.77
0.69
0.61
0.36
0.35
0.4
0.82
0.87
0.86
0.77
0.39
0.38
0.80
0.86
0.83
0.75
0.39
0.37
0.77
0.85
0.78
0.71
0.39
0.37
0.76
0.84
0.75
0.68
0.38
0.36
0.77
0.85
0.78
0.71
0.39
0.37
0.5
0.87
0.91
0.88
0.81
0.45
0.39
0.86
0.90
0.84
0.77
0.41
0.38
0.85
0.89
0.81
0.75
0.39
0.38
0.84
0.88
0.79
0.73
0.39
0.36
0.85
0.89
0.81
0.75
0.39
0.38
A-7
Table A-7 Standard fuzzy membership values p(x) for chip breakability Work Material: CS 1020 Chip Breaker: ENT-type Tool Geometry : 0°, 5°, -6°, 90°, 60°, 0.8
Cutting Speed (mAnin)
50
75
100
150
200
Depth of Cut (mm)
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
0.06
0.42
0.42
0.43
0.43
0.51
0.48
0.22
0.22
0.22
0.27
0.51
0.45
0.22
0.22
0.22
0.23
0.32
0.39
0.18
0.18
0.18
0.20
0.28
0.36
0.14
0.26
0.14
0.17
0.25
0.32
0.1
0.32
0.31
0.31
0.25
0.38
0.45
0.31
0.30
0.30
0.24
0.38
0.45
0.30
0.30
0.30
0.23
0.34
0.38
0.29
0.29
0.28
0.22
0.32
0.35
0.26
0.17
0.25
0.19
0.28
0.31
Feed (mm/rev)
0.2
0.63
0.62
0.61
0.51
0.39
0.38
0.61
0.60
0.59
0.47
0.37
0.38
0.59
0.58
0.57
0.44
0.33
0.36
0.57
0.57
0.55
0.42
0.30
0.36
0.53
0.37
0.51
0.38
0.26
0.33
0.3
0.79
0.80
0.77
0.70
0.40
0.38
0.77
0.78
0.75
0.68
0.39
0.38
0.75
0.76
0.73
0.66
0.38
0.36
0.73
0.74
0.71
0.62
0.37
0.36
0.69
0.70
0.67
0.58
0.33
0.33
0.4
0.89
0.91
0.82
0.79
0.43
0.40
0.87
0.89
0.80
0.74
0.42
0.40
0.85
0.87
0.78
0.70
0.41
0.38
0.83
0.84
0.76
0.66
0.40
0.37
0.79
0.80
0.72
0.62
0.36
0.34
0.5
0.92
0.95
0.88
0.81
0.45
0.43
0.91 !
0.94
0.86
0.78
0.44
0.43
0.90
0.93
0.84
0.75
0.43
0.41
0.89
0.92
0.82
0.76
0.42
0.41
0.85
0.88 j
0.78
0.68
0.38
0.40
A-8
Table A-8 Standard fuzzy membership values p(x) for chip breakability Work Material :K1040 Chip Breaker: ENT-type Tool Geometry : 0°, 5°, -6°, 90°, 60°, 0.8
Cutting Speed (rn/rnin)
50
75
100
150
200
Depth of Cut (mm)
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
0.06
0.42
0.42
0.43
0.43
0.55
0.55
0.20
0.20
0.20
0.25
0.35
0.40
0.20
0.20
0.20
0.21
0.30
0.37
0.15
0.15
0.15
0.18
0.26
0.34
0.12
0.12
0.12
0.15
0.23
0.30
0.1
0.46
0.40
0.40
0.40
0.38
0.46
0.30
0.30
0.22
0.28
0.34
0.38
0.25
0.25
0.22
0.22
0.32
0.36
0.20
0.20
0.18
0.18
0.30
0.33
0.15
0.15
0.14
0.15
0.26
0.30
Feed (mm/rev)
0.2
0.49
0.46
0.46
0.43
0.35
0.38
0.45
0.45
0.45
0.45
0.35
0.38
0.45
0.45
0.45
0.45
0.31
0.36
0.38
0.39
0.38
0.38
0.27
0.36
0.34
0.35
0.34
0.34
0.30
0.33
0.3
0.77
0.78
0.75
0.68
0.40
0.38
0.75
0.76
0.73
0.66
0.39
0.38
0.73
0.74
0.71
0.64
0.38
0.36
0.71
0.72
0.69
0.60
0.37
0.36
0.67
0.68
0.65
0.56
0.33
0.33
0.4
0.87
0.89 !
0.80
0.76
0.43
0.40
0.85
0.87
0.78
0.72 ;
0.42
0.40
0.83
0.85
0.76
0.68
0.41
0.38
0.81
0.82
0.74
0.64
0.40
0.37
0.77
0.78
0.70
0.60
0.36
0.34
0.5
0.90
0.93
0.86
0.79
0.45
0.43
0.89
0.92
0.84
0.76 1 0.44
0.43
0.88
0.91
0.82
0.73
0.43
0.41
0.87
0.90
0.80
0.70
0.42
0.41
0.83
0.86
0.76
0.66
0.38
I 0.39
A-9
Table A-9 Standard fuzzy membership values p(x) for chip breakability Work Material :EN25 Chip Breaker: ENT-type Tool Geometry : 0°, 5°, -6°, 90°, 60°, 0.8
Cutting Speed (rn/min)
50
75
100
150
200
Depth of Cut (mm)
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
4.0
3.0
2.0
1.0
0.5
0.25
Feed (mm/rev)
0.06
0.28
0.30
0.31
0.32
0.35
0.38
0.17
0.17
0.17
0.22
0.32
0.36
0.17
0.12
0.12
0.18
0.27
0.34
0.08
0.10
0.12
0.15
0.23
0.31
0.06
0.09
0.09
0.12
0.21
0.27
0.1
0.29
0.30
0.28
0.28
0.28
0.37
0.27
0.27
0.25
0.25
0.31
0.35
0.22
0.22
0.19
0.19
0.29
0.33
0.17
0.17
0.15
0.15
0.27
0.30
0.12
0.12
0.11
0.12
0.22
0.26
0.2
0.45
0.47
0.47
0.42
0.32
0.37
0.41
0.43
0.42
0.39
0.28
0.37
0.39
0.40
0.40
0.38
0.25
0.35
0.38
0.38
0.38
0.35
0.22
0.35
0.31
0.32
0.31
0.31
0.22
0.33
0.3
0.74
0.75
0.72
0.65
0.39
0.37
0.72
0.73
0.70
0.63
0.38
0.37
0.70
0.71
0.68
0.59
0.37
0.35
0.68
0.69
0.66
0.48
0.36
0.35
0.64
0.65
0.62
0.53
0.32
0.33
0.4
0.84
0.86
0.77
0.73
0.41
0.39
0.82
0.84
0.75
0.69
0.40
0.39
0.80
0.82
0.73
0.65
0.39
0.37
0.78
0.79
0.71
0.61
0.38
0.36
0.74
0.75
0.67
0.57
0.32
0.33
0.5
0.87
0.90
0.83
0.76
0.43
0.42
0.86
0.89
0.81
0.73
0.42
0.42
0.85
0.88
0.79
0.70
0.41
0.40
0.84
0.87
0.77
0.67
0.40
0.40
0.80
0.83
0.73
0.63
0.34
0.38
B-1
APPENDIX B
REPRESENTATIVE DATA OF SURFACE FINISH WHEN MACHINING WITH DIFFERENT TOOL CHIP BREAKERS
Table B-1 Surface roughness Ra (pm) for work material CS1020 Tool Geometry : 0°, 5°, -6°, 90°, 60°, 0.8 Depth of Cut: d = 0.5 m m Machine Tool: Colchester Mascot 1600 (9.3 K W )
Cutting Speed (rn/min)
100
150
200
Feed (mm/rev)
0.06
0.08
0.10
0.12
0.15
0.20
0.25
0.30
0.06
0.08
0.10
0.12
0.15
0.20
0.25
0.30
0.06
0.08
0.10
0.12
0.15
Tool Chi EFJ-type
1.69
1.38
1.22
1.30
1.37
1.63
2.42
3.54
1.15
1.17
1.20
1.24
1.30
1.62
2.29
3.22
1.04
1.10
1.16
1.20
1.25
EFB-type
1.38
1.29
1.17
1.20
1.38
1.66
2.38
3.41
1.08
1.10
1.15
1.17
1.33
1.48
2.28
3.01
0.95
0.99
1.04
1.12
1.18
ENA-type
1.28
1.22
1.29
1.32
1.41
1.84
2.57
3.87
1.17
1.20
1.26
1.31
1.37
1.77
2.19
3.37
1.08
1.13
1.17
1.24
1.31
) Breaker ENG- i type
1.27
1.20
1.26 |
1.32
1.60
1.99
2.59
3.87
1.12
1.18
1.23
1.26
1.50
1.88
2.31
3.19
1.07
1.12
1.13
1.18
1.29
ENK-type
1.95
1.61
1.29
1.38
1.62
1.82
2.61
3.79
1.21
1.23
1.28
1.32
1.64
1.81
2.09
3.20
1.11
1.15
1.18
1.25
1.31
ENJ-type
1.90
1.49
1.24
1.35
1.60
1.88
2.70
3.96
1.20
1.22
1.26
1.31
1.58
1.87
2.31
3.29
1.09
1.12
1.17
1.24
1.30
B-2
Table B-2 Surface roughness Ra (pm) for work material K1040 Tool Geometry : 0°, 5°, -6°, 90°, 60°, 0.8 Depth of Cut: d = 0.5 m m Machine Tool: Colchester Mascot 1600 (9.3 K W )
Cutting Speed (rrVmin)
100
150
200
Feed (mm/rev)
0.06
0.08
0.10
0.12
0.15
0.20
0.25
0.30
0.06
0.08
0.10
0.12
0.15
0.20
0.25
0.30
0.06
0.08
0.10
0.12
0.15
"Tool Chip Breaker EFJ-type
1.91
1.72
1.32
1.41
1.65
1.82
2.59
3.60
1.14
1.19
1.25
1.32
1.39
1.69
2.18
2.97
1.06
1.10
1.13
1.24
1.27
EFB-type
1.82
1.20
1.26
1.30
1.31
1.70
2.27
3.26
1.11
1.14
1.15
1.21
1.28
1.69
2.20
2.88
0.98
1.11
1.15
1.20
1.25
ENA-type
1.49
1.25
1.28
1.42
1.66
1.91
2.39
2.99
1.19
1.24
1.27
1.35
1.49
1.80
2.31
3.01
1.09
1.12
1.16
1.21
1.29
ENG-type
2.03
1.43
1.27
1.30
1.51
1.94
2.34
3.21
1.16
1.19
1.27
1.34
1.60
1.84
2.24
2.86
1.08
1.10
1.17
1.22
1.32
ENK-type
1.81
1.29
1.33
1.36
1.70
1.97
2.29
3.21
1.18
1.24
1.31
1.39
1.59
1.86
2.35
3.02
1.12
1.16
1.20
1.23
1.29
ENJ-type
2.19
1.80
1.31
1.38
1.59
1.96
2.34
3.39
1.15
1.22
1.30
1.37
1.61
1.91
2.44
3.26
1.10
1.13
1.17
1.24 i
1.29
Table B-3 Surface roughness Ra (pm) for work material EN25 Tool Geometry : 0°, 5°, -6°, 90°, 60°, 0.8 Depth of Cut: d = 0.5 m m Machine Tool: Colchester Mascot 1600 (9.3 K W )
Cutting Speed (rnAnin)
100
150
200
Feed (mm/rev)
0.06
0.08
0.10
0.12
0.15
0.20
0.25
0.30
0.06
0.08
0.10
0.12
0.15
0.20
0.25
0.30
0.06
0.08
0.10
0.12
0.15
Tool Chi EFJ- ! type
0.76
0.78
0.81
0.97
1.10
1.61
2.45
3.36
0.74
0.77
0.80
0.89
1.23
1.60
2.38
3.25
0.70
0.73
0.78
0.83
0.89
EFB-type
0.73
0.74
0.78
0.84
1.19
1.49
2.13
2.78
0.70
0.72
0.75
0.79
1.10
1.44
1.92
2.59
0.68
0.71
0.73
0.78
0.84
ENA-type
0.82
0.85
0.89
0.98
1.28
1.62
2.38
2.97
0.78
0.79
0.83
0.88
1.27
1.60
2.21
2.89
0.72
0.76
0.78
0.87
0.95
) Breaker ENG-type
0.80
0.84
0.86
0.95
1.27
1.71
2.58
3.11
0.78
0.81
0.84
0.87
1.19
1.71
2.32
2.88
0.73
0.74
0.82
0.86
0.96
ENK-type
0.89
0.92
0.97
1.01
1.29
1.80
2.49
2.88
0.84
0.88
0.91
0.93
1.34
1.85
2.25
2.91
0.78
0.83
0.88
0.91
1.04
ENJ-type
0.84
0.87
0.91
0.99
1.32 j
1.88
2.61
3.24
0.82
0.83
0.87
0.92
1.31
1.89
2.61
3.11
0.76
0.79
0.84
0.89
1.08
C-1
APPENDIX C
OVERALL GROOVE WEAR DATA FROM GROOVE WEAR EXPERIMENTS
Table C-1 Overall groove wear data under cutting condition Group A Machine Tool: HITEC-20SII CNC Lathe (18 K W ) V = 160 m/min, f = 0.08 mm/rev, d = 0.25 m m
Cutting
Time
(minute)
0.0
0.5
1.0
1.5
2.0
3.0
4.0
5.0
6.0
7.5
8.5
10
12
14
16
18
20
22
25
27
30
33
35.5
38
41
45
48
Number
of
Groove
0
2
3
3
3
4
4
5
5
Groove
Depth
(pm)
0
5
5.5
6
6.5
6.9
8.2
9.3
10.7
Maximum Groove Length (pm)
0
91
105
140
155
165
185
225
260
Groove Wear Area (mm2)
0
0.49
0.84
2.10
2.39
3.51
4.25
7.60
7.80
Nose
wear
(pm)
0
4
5
6
8
10
13
17
20
Surface
Roughness
Ra(pm)
0.86
0.92
0.95
1.29
1.59
1.68
1.74
1.77
1.79
1.78
1.80
1.81
1.79
1.82
1.80
1.77
1.77
1.78
1.99
2.23
2.41
2.70
2.89
2.98
3.25
3.29
3.37
C-2
Table C-2 Overall groove wear data under cutting condition Group B Machine Tool: HJTEC-20SII CNC Lathe (18 KW) V = 160 m/min, f = 0.04 mnvrev, d = 0.25 m m
Cutting
Time
(minute)
0.0
0.5
1.0
1.5
2.0
3.0
4.0
5.0
6.0
7.5
8.5
10
12
14
16
18.5
21
23
25
27.5
30
33
37
40
43
47
50
Number
of
Groove
0
4
5
7
8
4
5
5
1
Groove
Depth
(pm)
0
5.0
5.0
5.5
5.8
4.7
6.8
5.3
5.2
Maximum Groove Length (pm)
o 1
65
113
150
160
175
225
250
265
Groove Wear Area (mm2)
0
0.52
1.31
3.45
4.20
4.26
6.53
7.50
8.32
Nose
wear
(um)
0
4.5
6
7
8.5
12
15
18
21
Surface
Roughness
Ra(um)
0.40
0.44
0.47
0.70
0.98
1.36
1.58
1.71
1.79
1.82
1.85
1.82
1.82
1.80
1.84
2.98
3.51
3.40
3.30
3.20
3.31
3.22
3.13
3.14
3.50
3.55
3.58
C-3
Table C-3 Overall groove wear data under cutting condition Group C Machine Tool: HJTEC-20Sn C N C Lathe (18 K W ) V = 125 m/min, f = 0.08 mirvrcv, d = 0.25 m m
Cutting
Time
(minute)
0.0
0.5
1.0
1.5
2.5
3.5
5.0
7.0
9.0
11
13
15
17
19
21
23
26
29
32
35
39
43
46
49
53
57
61
Number
of
Groove
0
2
2
3
3
3
4
5
5
Groove
Depth
(um)
0
5.0
6.0
6.5
6.7
6.7
6.8
10.3
10.5
Maximum Groove Length (pm)
0
67
82
135
158
165
205
310
330
Groove Wear Area (mm2) 0
0.35
0.60
2.06
2.56
2.72
4.40
4.62
10.5
Nose
wear
(pm)
0
4.5
5
7
9
12.5
15
19
23
Surface
Roughness
Ra(um)
1.02
1.02
1.05
1.40
1.68
1.79
1.80
1.83
1.82 |
1.83
1.82
1.82
1.83
1.84
1.87
1.84
1.85
1.86
1.83
1.79
1.79
1.91
2.38
2.77
2.98
3.19
3.28
C-4
Table C-4 Overall groove wear data under cutting condition Group D Machine Tool: HrTEC-20SJJ. C N C Lathe (18 K W ) V = 190 m/min, f = 0.08mm£ev, d = 0.25 m m
Cutting
lime
(minute)
0.0
0.5 |
1.0
1.5
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9
10.5
12
14
16
18
20
22
25
27
29
31
34.5
37
39.5
42
Number
of
Groove
0
2
3
3
3
3
4
4
5
Groove
Depth
(pm)
0
4.5
4.5
4.5
5.0
5.3
7.3
8.7
9.6
Maximum Groove Length (pm)
0
72
85
112
135
160
180
220
280
Groove Wear Area (mm2)
0
0.38
0.85
1.66
2.30
2.50
3.96
4.90
8.75
Nose
wear
(pm)
0
6
7
8.5
10
11.5
17.5
20
24
Surface 1
Roughness
Ra(um)
0.79
0.82
0.87
1.17
1.42
1.58
1.65
1.69
1.70
1.72
1.72
1.73
1.74
1.76
1.77
1.79
1.80
1.82
1.85
1.89
1.94
2.09
2.23
2.49
2.68
2.89
3.19
C-5
Table C-5 Overall groove wear data under cutting condition Group E Machine Tool: HTrEC-20Sn CNC Lathe (18 K W ) V = 190 m/min, f = 0.04 mirvTev, d = 0.25 m m
Cutting
Time
(minute)
0.0
0.5
1.0
1.5
2.0
3.0
4.0
5.0
6.0
7.5
9
11
13
15
17
18.5
21
23
25
27
30
32
35
37
39
41
44
Number
of
Groove
0
4
5
7
7
2
1
1
1
Groove
Depth
(pm)
0
4.5
4.5
4.8
6.1
4.3
2.8
2.65
2.30
Maximum Groove Length (pm)
0
75
95
145
160
200
240
275
318
Groove Wear Area (mm2)
0 j
0.94
1.12
3.22
3.68
6.24
6.48
7.80
9.99
Nose
wear
(pm)
0
6
7
7.5
10
15
18
21
25
Surface
Roughness
Ra(um)
0.35
0.37
0.40
0.70
0.99
1.34
1.56
1.74
1.73
1.76
1.79
1.81
1.86
1.85
1.84
2.59
2.99
3.51
3.62
3.69
3.62
3.57
3.58
3.60
3.59
3.58
3.60
D-1
APPENDIX D
INPUT/OUTPUT TRAINING AND TESTING DATA IN NEURAL NETWORK EXPERIMENTS
Table D-1 The training data under Training Cutting Conditions 4-6
Tf
i 6 •i 3
u
IO 3
1 8 00
•s +->
3
u NO
§ •43
•-a c
a 3
1 u
INPUT FEATURES
Feed Speed D (LF) D£(HF) DJ(HF) D£(HF)
(mm/rev) (m/min) (min" ) (min"1) (min"1) (min"1)
0.06
0.06
0.06
0.06
0.06
0.06
0.10
0.10
0.10
0.10
0.10
0.10
0.10
0.15
0.15
0.15
0.15
0.15
0.15
205
205
205
205
205
205
170
170
170
170
170
170
170
160
160
160
160
160
160
-8.09
-4.57
-1 .05
2.47
5.99
9.51
-4.82
-2.82
-0.82
-1.18
3.18
5.18
7.18
-3.65
-1.17
1.31
3.79
6.27
8.75
2.07
1.23
0.39
-0.45
-1.29
-2.13
3.16
2.06
0.96
-0.14
-1 .24
-2.34
-3.44
3.32
1.66
0.00
-1 .66
-3.32
-4.98
4.53
2.38
0.23
-1 .93
-4.08
-8.59
3.44
2.09
0.74
-0.61
-1 .96
-3.31
-4.66
6.75
4.83
2.91
0.99
-0.93
-2.85
2.17
1.35
0.53
-0.29
-1.11
-1.93
0.53
0.18
-0.17
-0.52
-0.87
-1.22
-1.57
1.31
0.87
0.43
0.01
-0.45
-0.89
0.15
0.15
0.20
0.28
0.35
0.30
0.22
0.22
0.25
0.29
0.33
0.38
0.37
0.28
0.28
0.35
0.48
0.51
0.45
OUTPUTS
Ra ^0(k) (pm)
0.15
0.20
0.28
0.35
0.30
0.26
0.22
0.25
0.29
0.33
0.38
0.37
0.34
0.28
0.35
0.48
0.51
0.45
0.42
0.85
1.00
1.12
1.49
1.74
1.99
0.94
1.15
1.35
1.50
1.68
2.09
2.27
0.98
1.14
1.22
1.39
1.81
1.96
D
Table D-2 The testing data under Testing Cutting Conditions 1-2
3
•S
1 a w>
•n
cs 3 •B
-a
a on 3 •43 Vi
#
INPUT FEATURES
Feed Speed Dx(LF) D (HF) IX(HF) D fllF) 1 l l i M" i (k)
(mm/rev) (m/min) (min ) (min ) (min ) (min)
0.12
0.12
0.12
0.12
0.12
0.12
0.12
0.06
0.06
0.06
0.06
0.06
0.06
0.06
140
140
140
140
140
140
140
165
165
165
165
165
165
165
-9.78
-6.51
-3.23
0.05
3.32
6.60
9.87
-0.97
-6.22
-3.37
-0.52
2.33
5.18
8.03
6.01
4.14
2.26
0.39
-1 .49
-3.37
-5.24
3.44
2.34
1.24
0.14
-0.96
-2.06
-3.16
2.55
1.40
0.25
-0.90
-2.05
-3.20
-4.35
6.03
4.03
2.03
0.03
-1.97
-3.97
-5.97
1.52
1.15
0.77
0.40
0.02
-0.36
-0.73
1.26
0.86
0.46
0.06
-0.34
-0.74
-1.14
0.28
0.28
0.31
0.34
0.38
0.53
0.50
0.20
0.20
0.23
0.26
0.30
0.36
0.38
OUTPUTS
Ra ^0(R) (pm)
0.28
0.31
0.34
0.38
0.53
0.50
0.44
0.20
0.23
0.26
0.30
0.36
0.38
0.34
1.06
1.09
1.13
1.33
1.87
2.05
2.33
0.98
1.08
1.12
1.26
1.52
1.94
2.37
D
Table D-3 The testing data under Testing Cutting Conditions 3-4
cn
1 8
6 W) 3 •43 Vi
Tf 3
•S
1 a Wi
•a
INPUT FEATURES
Feed Speed D;(LF) D^(HF) D£(HF) D^(HF)
(mm/'rev) (m/min) (min'1) (min"1) (min"1) (min"1)
0.06
0.06
0.06
0.06
0.06
0.06
0.15
0.15
0.15
0.15
0.15
0.15
0.15
0.15
190
190
190
190
190
190
130
130
130
130
130
130
130
130
-1 2.9
-9.00
-5.12
-1 .24
2.64
6.53
-4.54
-3.02
-1 .49
0.04
1.56
3.09
4.61
6.44
1.23
0.58
-0.07
-0.80
-1.37
-2.02
3.87
2.72
1.57
0.42
-0.73
-1.88
-3.03
-4.41
8.59
6.31
4.03
1.75
-0.53
-2.81
6.69
4.64
2.59
0.54
-1.51
-3.56
-5.61
-8.07
0.63
0.37
0.10
-0.17
-0.43
-0.70
2.38
1.78
1.18
0.58
-0.02
-0.62
-1.22
-1.94
0.18
0.18
0.22
0.26
0.33
0.36
0.30
0.30
0.32
0.37
0.45
0.52
0.56
0.51
OUTPUTS
Ra ^0( k ) (pm)
0.18
0.22
0.26
0.33
0.36
0.31
0.30
0.32
0.37
0.45
0.52
0.56
0.51
0.49
0.95
1.11
1.21
1.37
1.58
1.92
1.10
1.13
1.16
1.22
1.34
1.67
1.96
2.23
![ResearchonShapeChangesinCylinderElectrodes IncidenttoMicro-EDMdownloads.hindawi.com/journals/amse/2019/8087462.pdf · micro-EDM milling. e electrode wear estimation model [5]isusedfor](https://static.fdocuments.net/doc/165x107/5e7374c5941139347653a813/researchonshapechangesincylinderelectrodes-incidenttomicro-micro-edm-milling-e.jpg)
